spi-topcliff-pch: add recovery processing in case wait-event timeout
[zen-stable.git] / net / ipv6 / addrconf.c
blob6b8ebc5da0e1344ad58994e327ba42cac44f3ac3
1 /*
2 * IPv6 Address [auto]configuration
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
16 * Changes:
18 * Janos Farkas : delete timer on ifdown
19 * <chexum@bankinf.banki.hu>
20 * Andi Kleen : kill double kfree on module
21 * unload.
22 * Maciej W. Rozycki : FDDI support
23 * sekiya@USAGI : Don't send too many RS
24 * packets.
25 * yoshfuji@USAGI : Fixed interval between DAD
26 * packets.
27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of
28 * address validation timer.
29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041)
30 * support.
31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6
32 * address on a same interface.
33 * YOSHIFUJI Hideaki @USAGI : ARCnet support
34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to
35 * seq_file.
36 * YOSHIFUJI Hideaki @USAGI : improved source address
37 * selection; consider scope,
38 * status etc.
41 #include <linux/errno.h>
42 #include <linux/types.h>
43 #include <linux/kernel.h>
44 #include <linux/socket.h>
45 #include <linux/sockios.h>
46 #include <linux/net.h>
47 #include <linux/in6.h>
48 #include <linux/netdevice.h>
49 #include <linux/if_addr.h>
50 #include <linux/if_arp.h>
51 #include <linux/if_arcnet.h>
52 #include <linux/if_infiniband.h>
53 #include <linux/route.h>
54 #include <linux/inetdevice.h>
55 #include <linux/init.h>
56 #include <linux/slab.h>
57 #ifdef CONFIG_SYSCTL
58 #include <linux/sysctl.h>
59 #endif
60 #include <linux/capability.h>
61 #include <linux/delay.h>
62 #include <linux/notifier.h>
63 #include <linux/string.h>
65 #include <net/net_namespace.h>
66 #include <net/sock.h>
67 #include <net/snmp.h>
69 #include <net/ipv6.h>
70 #include <net/protocol.h>
71 #include <net/ndisc.h>
72 #include <net/ip6_route.h>
73 #include <net/addrconf.h>
74 #include <net/tcp.h>
75 #include <net/ip.h>
76 #include <net/netlink.h>
77 #include <net/pkt_sched.h>
78 #include <linux/if_tunnel.h>
79 #include <linux/rtnetlink.h>
81 #ifdef CONFIG_IPV6_PRIVACY
82 #include <linux/random.h>
83 #endif
85 #include <linux/uaccess.h>
86 #include <asm/unaligned.h>
88 #include <linux/proc_fs.h>
89 #include <linux/seq_file.h>
90 #include <linux/export.h>
92 /* Set to 3 to get tracing... */
93 #define ACONF_DEBUG 2
95 #if ACONF_DEBUG >= 3
96 #define ADBG(x) printk x
97 #else
98 #define ADBG(x)
99 #endif
101 #define INFINITY_LIFE_TIME 0xFFFFFFFF
103 static inline u32 cstamp_delta(unsigned long cstamp)
105 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
108 #define ADDRCONF_TIMER_FUZZ_MINUS (HZ > 50 ? HZ/50 : 1)
109 #define ADDRCONF_TIMER_FUZZ (HZ / 4)
110 #define ADDRCONF_TIMER_FUZZ_MAX (HZ)
112 #ifdef CONFIG_SYSCTL
113 static void addrconf_sysctl_register(struct inet6_dev *idev);
114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
115 #else
116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
123 #endif
125 #ifdef CONFIG_IPV6_PRIVACY
126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
128 static void ipv6_regen_rndid(unsigned long data);
129 #endif
131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
132 static int ipv6_count_addresses(struct inet6_dev *idev);
135 * Configured unicast address hash table
137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
138 static DEFINE_SPINLOCK(addrconf_hash_lock);
140 static void addrconf_verify(unsigned long);
142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
143 static DEFINE_SPINLOCK(addrconf_verify_lock);
145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
148 static void addrconf_type_change(struct net_device *dev,
149 unsigned long event);
150 static int addrconf_ifdown(struct net_device *dev, int how);
152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
153 static void addrconf_dad_timer(unsigned long data);
154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
155 static void addrconf_dad_run(struct inet6_dev *idev);
156 static void addrconf_rs_timer(unsigned long data);
157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
161 struct prefix_info *pinfo);
162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
163 struct net_device *dev);
165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
168 .forwarding = 0,
169 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
170 .mtu6 = IPV6_MIN_MTU,
171 .accept_ra = 1,
172 .accept_redirects = 1,
173 .autoconf = 1,
174 .force_mld_version = 0,
175 .dad_transmits = 1,
176 .rtr_solicits = MAX_RTR_SOLICITATIONS,
177 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
178 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
179 #ifdef CONFIG_IPV6_PRIVACY
180 .use_tempaddr = 0,
181 .temp_valid_lft = TEMP_VALID_LIFETIME,
182 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
183 .regen_max_retry = REGEN_MAX_RETRY,
184 .max_desync_factor = MAX_DESYNC_FACTOR,
185 #endif
186 .max_addresses = IPV6_MAX_ADDRESSES,
187 .accept_ra_defrtr = 1,
188 .accept_ra_pinfo = 1,
189 #ifdef CONFIG_IPV6_ROUTER_PREF
190 .accept_ra_rtr_pref = 1,
191 .rtr_probe_interval = 60 * HZ,
192 #ifdef CONFIG_IPV6_ROUTE_INFO
193 .accept_ra_rt_info_max_plen = 0,
194 #endif
195 #endif
196 .proxy_ndp = 0,
197 .accept_source_route = 0, /* we do not accept RH0 by default. */
198 .disable_ipv6 = 0,
199 .accept_dad = 1,
202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
203 .forwarding = 0,
204 .hop_limit = IPV6_DEFAULT_HOPLIMIT,
205 .mtu6 = IPV6_MIN_MTU,
206 .accept_ra = 1,
207 .accept_redirects = 1,
208 .autoconf = 1,
209 .dad_transmits = 1,
210 .rtr_solicits = MAX_RTR_SOLICITATIONS,
211 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL,
212 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY,
213 #ifdef CONFIG_IPV6_PRIVACY
214 .use_tempaddr = 0,
215 .temp_valid_lft = TEMP_VALID_LIFETIME,
216 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME,
217 .regen_max_retry = REGEN_MAX_RETRY,
218 .max_desync_factor = MAX_DESYNC_FACTOR,
219 #endif
220 .max_addresses = IPV6_MAX_ADDRESSES,
221 .accept_ra_defrtr = 1,
222 .accept_ra_pinfo = 1,
223 #ifdef CONFIG_IPV6_ROUTER_PREF
224 .accept_ra_rtr_pref = 1,
225 .rtr_probe_interval = 60 * HZ,
226 #ifdef CONFIG_IPV6_ROUTE_INFO
227 .accept_ra_rt_info_max_plen = 0,
228 #endif
229 #endif
230 .proxy_ndp = 0,
231 .accept_source_route = 0, /* we do not accept RH0 by default. */
232 .disable_ipv6 = 0,
233 .accept_dad = 1,
236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
242 /* Check if a valid qdisc is available */
243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
245 return !qdisc_tx_is_noop(dev);
248 /* Check if a route is valid prefix route */
249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
251 return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
256 if (del_timer(&ifp->timer))
257 __in6_ifa_put(ifp);
260 enum addrconf_timer_t {
261 AC_NONE,
262 AC_DAD,
263 AC_RS,
266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
267 enum addrconf_timer_t what,
268 unsigned long when)
270 if (!del_timer(&ifp->timer))
271 in6_ifa_hold(ifp);
273 switch (what) {
274 case AC_DAD:
275 ifp->timer.function = addrconf_dad_timer;
276 break;
277 case AC_RS:
278 ifp->timer.function = addrconf_rs_timer;
279 break;
280 default:
281 break;
283 ifp->timer.expires = jiffies + when;
284 add_timer(&ifp->timer);
287 static int snmp6_alloc_dev(struct inet6_dev *idev)
289 if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
290 sizeof(struct ipstats_mib),
291 __alignof__(struct ipstats_mib)) < 0)
292 goto err_ip;
293 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
294 GFP_KERNEL);
295 if (!idev->stats.icmpv6dev)
296 goto err_icmp;
297 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
298 GFP_KERNEL);
299 if (!idev->stats.icmpv6msgdev)
300 goto err_icmpmsg;
302 return 0;
304 err_icmpmsg:
305 kfree(idev->stats.icmpv6dev);
306 err_icmp:
307 snmp_mib_free((void __percpu **)idev->stats.ipv6);
308 err_ip:
309 return -ENOMEM;
312 static void snmp6_free_dev(struct inet6_dev *idev)
314 kfree(idev->stats.icmpv6msgdev);
315 kfree(idev->stats.icmpv6dev);
316 snmp_mib_free((void __percpu **)idev->stats.ipv6);
319 /* Nobody refers to this device, we may destroy it. */
321 void in6_dev_finish_destroy(struct inet6_dev *idev)
323 struct net_device *dev = idev->dev;
325 WARN_ON(!list_empty(&idev->addr_list));
326 WARN_ON(idev->mc_list != NULL);
328 #ifdef NET_REFCNT_DEBUG
329 printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
330 #endif
331 dev_put(dev);
332 if (!idev->dead) {
333 pr_warning("Freeing alive inet6 device %p\n", idev);
334 return;
336 snmp6_free_dev(idev);
337 kfree_rcu(idev, rcu);
340 EXPORT_SYMBOL(in6_dev_finish_destroy);
342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
344 struct inet6_dev *ndev;
346 ASSERT_RTNL();
348 if (dev->mtu < IPV6_MIN_MTU)
349 return NULL;
351 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
353 if (ndev == NULL)
354 return NULL;
356 rwlock_init(&ndev->lock);
357 ndev->dev = dev;
358 INIT_LIST_HEAD(&ndev->addr_list);
360 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
361 ndev->cnf.mtu6 = dev->mtu;
362 ndev->cnf.sysctl = NULL;
363 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
364 if (ndev->nd_parms == NULL) {
365 kfree(ndev);
366 return NULL;
368 if (ndev->cnf.forwarding)
369 dev_disable_lro(dev);
370 /* We refer to the device */
371 dev_hold(dev);
373 if (snmp6_alloc_dev(ndev) < 0) {
374 ADBG((KERN_WARNING
375 "%s(): cannot allocate memory for statistics; dev=%s.\n",
376 __func__, dev->name));
377 neigh_parms_release(&nd_tbl, ndev->nd_parms);
378 dev_put(dev);
379 kfree(ndev);
380 return NULL;
383 if (snmp6_register_dev(ndev) < 0) {
384 ADBG((KERN_WARNING
385 "%s(): cannot create /proc/net/dev_snmp6/%s\n",
386 __func__, dev->name));
387 neigh_parms_release(&nd_tbl, ndev->nd_parms);
388 ndev->dead = 1;
389 in6_dev_finish_destroy(ndev);
390 return NULL;
393 /* One reference from device. We must do this before
394 * we invoke __ipv6_regen_rndid().
396 in6_dev_hold(ndev);
398 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
399 ndev->cnf.accept_dad = -1;
401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
402 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
403 printk(KERN_INFO
404 "%s: Disabled Multicast RS\n",
405 dev->name);
406 ndev->cnf.rtr_solicits = 0;
408 #endif
410 #ifdef CONFIG_IPV6_PRIVACY
411 INIT_LIST_HEAD(&ndev->tempaddr_list);
412 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
413 if ((dev->flags&IFF_LOOPBACK) ||
414 dev->type == ARPHRD_TUNNEL ||
415 dev->type == ARPHRD_TUNNEL6 ||
416 dev->type == ARPHRD_SIT ||
417 dev->type == ARPHRD_NONE) {
418 ndev->cnf.use_tempaddr = -1;
419 } else {
420 in6_dev_hold(ndev);
421 ipv6_regen_rndid((unsigned long) ndev);
423 #endif
425 if (netif_running(dev) && addrconf_qdisc_ok(dev))
426 ndev->if_flags |= IF_READY;
428 ipv6_mc_init_dev(ndev);
429 ndev->tstamp = jiffies;
430 addrconf_sysctl_register(ndev);
431 /* protected by rtnl_lock */
432 rcu_assign_pointer(dev->ip6_ptr, ndev);
434 /* Join all-node multicast group */
435 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
437 /* Join all-router multicast group if forwarding is set */
438 if (ndev->cnf.forwarding && dev && (dev->flags & IFF_MULTICAST))
439 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
441 return ndev;
444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
446 struct inet6_dev *idev;
448 ASSERT_RTNL();
450 idev = __in6_dev_get(dev);
451 if (!idev) {
452 idev = ipv6_add_dev(dev);
453 if (!idev)
454 return NULL;
457 if (dev->flags&IFF_UP)
458 ipv6_mc_up(idev);
459 return idev;
462 #ifdef CONFIG_SYSCTL
463 static void dev_forward_change(struct inet6_dev *idev)
465 struct net_device *dev;
466 struct inet6_ifaddr *ifa;
468 if (!idev)
469 return;
470 dev = idev->dev;
471 if (idev->cnf.forwarding)
472 dev_disable_lro(dev);
473 if (dev && (dev->flags & IFF_MULTICAST)) {
474 if (idev->cnf.forwarding)
475 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
476 else
477 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
480 list_for_each_entry(ifa, &idev->addr_list, if_list) {
481 if (ifa->flags&IFA_F_TENTATIVE)
482 continue;
483 if (idev->cnf.forwarding)
484 addrconf_join_anycast(ifa);
485 else
486 addrconf_leave_anycast(ifa);
491 static void addrconf_forward_change(struct net *net, __s32 newf)
493 struct net_device *dev;
494 struct inet6_dev *idev;
496 rcu_read_lock();
497 for_each_netdev_rcu(net, dev) {
498 idev = __in6_dev_get(dev);
499 if (idev) {
500 int changed = (!idev->cnf.forwarding) ^ (!newf);
501 idev->cnf.forwarding = newf;
502 if (changed)
503 dev_forward_change(idev);
506 rcu_read_unlock();
509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
511 struct net *net;
512 int old;
514 if (!rtnl_trylock())
515 return restart_syscall();
517 net = (struct net *)table->extra2;
518 old = *p;
519 *p = newf;
521 if (p == &net->ipv6.devconf_dflt->forwarding) {
522 rtnl_unlock();
523 return 0;
526 if (p == &net->ipv6.devconf_all->forwarding) {
527 net->ipv6.devconf_dflt->forwarding = newf;
528 addrconf_forward_change(net, newf);
529 } else if ((!newf) ^ (!old))
530 dev_forward_change((struct inet6_dev *)table->extra1);
531 rtnl_unlock();
533 if (newf)
534 rt6_purge_dflt_routers(net);
535 return 1;
537 #endif
539 /* Nobody refers to this ifaddr, destroy it */
540 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
542 WARN_ON(!hlist_unhashed(&ifp->addr_lst));
544 #ifdef NET_REFCNT_DEBUG
545 printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
546 #endif
548 in6_dev_put(ifp->idev);
550 if (del_timer(&ifp->timer))
551 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
553 if (ifp->state != INET6_IFADDR_STATE_DEAD) {
554 pr_warning("Freeing alive inet6 address %p\n", ifp);
555 return;
557 dst_release(&ifp->rt->dst);
559 kfree_rcu(ifp, rcu);
562 static void
563 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
565 struct list_head *p;
566 int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
569 * Each device address list is sorted in order of scope -
570 * global before linklocal.
572 list_for_each(p, &idev->addr_list) {
573 struct inet6_ifaddr *ifa
574 = list_entry(p, struct inet6_ifaddr, if_list);
575 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
576 break;
579 list_add_tail(&ifp->if_list, p);
582 static u32 ipv6_addr_hash(const struct in6_addr *addr)
585 * We perform the hash function over the last 64 bits of the address
586 * This will include the IEEE address token on links that support it.
588 return jhash_2words((__force u32)addr->s6_addr32[2],
589 (__force u32)addr->s6_addr32[3], 0)
590 & (IN6_ADDR_HSIZE - 1);
593 /* On success it returns ifp with increased reference count */
595 static struct inet6_ifaddr *
596 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
597 int scope, u32 flags)
599 struct inet6_ifaddr *ifa = NULL;
600 struct rt6_info *rt;
601 unsigned int hash;
602 int err = 0;
603 int addr_type = ipv6_addr_type(addr);
605 if (addr_type == IPV6_ADDR_ANY ||
606 addr_type & IPV6_ADDR_MULTICAST ||
607 (!(idev->dev->flags & IFF_LOOPBACK) &&
608 addr_type & IPV6_ADDR_LOOPBACK))
609 return ERR_PTR(-EADDRNOTAVAIL);
611 rcu_read_lock_bh();
612 if (idev->dead) {
613 err = -ENODEV; /*XXX*/
614 goto out2;
617 if (idev->cnf.disable_ipv6) {
618 err = -EACCES;
619 goto out2;
622 spin_lock(&addrconf_hash_lock);
624 /* Ignore adding duplicate addresses on an interface */
625 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
626 ADBG(("ipv6_add_addr: already assigned\n"));
627 err = -EEXIST;
628 goto out;
631 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
633 if (ifa == NULL) {
634 ADBG(("ipv6_add_addr: malloc failed\n"));
635 err = -ENOBUFS;
636 goto out;
639 rt = addrconf_dst_alloc(idev, addr, false);
640 if (IS_ERR(rt)) {
641 err = PTR_ERR(rt);
642 goto out;
645 ifa->addr = *addr;
647 spin_lock_init(&ifa->lock);
648 spin_lock_init(&ifa->state_lock);
649 init_timer(&ifa->timer);
650 INIT_HLIST_NODE(&ifa->addr_lst);
651 ifa->timer.data = (unsigned long) ifa;
652 ifa->scope = scope;
653 ifa->prefix_len = pfxlen;
654 ifa->flags = flags | IFA_F_TENTATIVE;
655 ifa->cstamp = ifa->tstamp = jiffies;
657 ifa->rt = rt;
659 ifa->idev = idev;
660 in6_dev_hold(idev);
661 /* For caller */
662 in6_ifa_hold(ifa);
664 /* Add to big hash table */
665 hash = ipv6_addr_hash(addr);
667 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
668 spin_unlock(&addrconf_hash_lock);
670 write_lock(&idev->lock);
671 /* Add to inet6_dev unicast addr list. */
672 ipv6_link_dev_addr(idev, ifa);
674 #ifdef CONFIG_IPV6_PRIVACY
675 if (ifa->flags&IFA_F_TEMPORARY) {
676 list_add(&ifa->tmp_list, &idev->tempaddr_list);
677 in6_ifa_hold(ifa);
679 #endif
681 in6_ifa_hold(ifa);
682 write_unlock(&idev->lock);
683 out2:
684 rcu_read_unlock_bh();
686 if (likely(err == 0))
687 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
688 else {
689 kfree(ifa);
690 ifa = ERR_PTR(err);
693 return ifa;
694 out:
695 spin_unlock(&addrconf_hash_lock);
696 goto out2;
699 /* This function wants to get referenced ifp and releases it before return */
701 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
703 struct inet6_ifaddr *ifa, *ifn;
704 struct inet6_dev *idev = ifp->idev;
705 int state;
706 int deleted = 0, onlink = 0;
707 unsigned long expires = jiffies;
709 spin_lock_bh(&ifp->state_lock);
710 state = ifp->state;
711 ifp->state = INET6_IFADDR_STATE_DEAD;
712 spin_unlock_bh(&ifp->state_lock);
714 if (state == INET6_IFADDR_STATE_DEAD)
715 goto out;
717 spin_lock_bh(&addrconf_hash_lock);
718 hlist_del_init_rcu(&ifp->addr_lst);
719 spin_unlock_bh(&addrconf_hash_lock);
721 write_lock_bh(&idev->lock);
722 #ifdef CONFIG_IPV6_PRIVACY
723 if (ifp->flags&IFA_F_TEMPORARY) {
724 list_del(&ifp->tmp_list);
725 if (ifp->ifpub) {
726 in6_ifa_put(ifp->ifpub);
727 ifp->ifpub = NULL;
729 __in6_ifa_put(ifp);
731 #endif
733 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
734 if (ifa == ifp) {
735 list_del_init(&ifp->if_list);
736 __in6_ifa_put(ifp);
738 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
739 break;
740 deleted = 1;
741 continue;
742 } else if (ifp->flags & IFA_F_PERMANENT) {
743 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
744 ifp->prefix_len)) {
745 if (ifa->flags & IFA_F_PERMANENT) {
746 onlink = 1;
747 if (deleted)
748 break;
749 } else {
750 unsigned long lifetime;
752 if (!onlink)
753 onlink = -1;
755 spin_lock(&ifa->lock);
757 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
759 * Note: Because this address is
760 * not permanent, lifetime <
761 * LONG_MAX / HZ here.
763 if (time_before(expires,
764 ifa->tstamp + lifetime * HZ))
765 expires = ifa->tstamp + lifetime * HZ;
766 spin_unlock(&ifa->lock);
771 write_unlock_bh(&idev->lock);
773 addrconf_del_timer(ifp);
775 ipv6_ifa_notify(RTM_DELADDR, ifp);
777 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
780 * Purge or update corresponding prefix
782 * 1) we don't purge prefix here if address was not permanent.
783 * prefix is managed by its own lifetime.
784 * 2) if there're no addresses, delete prefix.
785 * 3) if there're still other permanent address(es),
786 * corresponding prefix is still permanent.
787 * 4) otherwise, update prefix lifetime to the
788 * longest valid lifetime among the corresponding
789 * addresses on the device.
790 * Note: subsequent RA will update lifetime.
792 * --yoshfuji
794 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
795 struct in6_addr prefix;
796 struct rt6_info *rt;
797 struct net *net = dev_net(ifp->idev->dev);
798 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
799 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
801 if (rt && addrconf_is_prefix_route(rt)) {
802 if (onlink == 0) {
803 ip6_del_rt(rt);
804 rt = NULL;
805 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
806 rt->dst.expires = expires;
807 rt->rt6i_flags |= RTF_EXPIRES;
810 dst_release(&rt->dst);
813 /* clean up prefsrc entries */
814 rt6_remove_prefsrc(ifp);
815 out:
816 in6_ifa_put(ifp);
819 #ifdef CONFIG_IPV6_PRIVACY
820 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
822 struct inet6_dev *idev = ifp->idev;
823 struct in6_addr addr, *tmpaddr;
824 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
825 unsigned long regen_advance;
826 int tmp_plen;
827 int ret = 0;
828 int max_addresses;
829 u32 addr_flags;
830 unsigned long now = jiffies;
832 write_lock(&idev->lock);
833 if (ift) {
834 spin_lock_bh(&ift->lock);
835 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
836 spin_unlock_bh(&ift->lock);
837 tmpaddr = &addr;
838 } else {
839 tmpaddr = NULL;
841 retry:
842 in6_dev_hold(idev);
843 if (idev->cnf.use_tempaddr <= 0) {
844 write_unlock(&idev->lock);
845 printk(KERN_INFO
846 "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
847 in6_dev_put(idev);
848 ret = -1;
849 goto out;
851 spin_lock_bh(&ifp->lock);
852 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
853 idev->cnf.use_tempaddr = -1; /*XXX*/
854 spin_unlock_bh(&ifp->lock);
855 write_unlock(&idev->lock);
856 printk(KERN_WARNING
857 "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
858 in6_dev_put(idev);
859 ret = -1;
860 goto out;
862 in6_ifa_hold(ifp);
863 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
864 if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
865 spin_unlock_bh(&ifp->lock);
866 write_unlock(&idev->lock);
867 printk(KERN_WARNING
868 "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
869 in6_ifa_put(ifp);
870 in6_dev_put(idev);
871 ret = -1;
872 goto out;
874 memcpy(&addr.s6_addr[8], idev->rndid, 8);
875 age = (now - ifp->tstamp) / HZ;
876 tmp_valid_lft = min_t(__u32,
877 ifp->valid_lft,
878 idev->cnf.temp_valid_lft + age);
879 tmp_prefered_lft = min_t(__u32,
880 ifp->prefered_lft,
881 idev->cnf.temp_prefered_lft + age -
882 idev->cnf.max_desync_factor);
883 tmp_plen = ifp->prefix_len;
884 max_addresses = idev->cnf.max_addresses;
885 tmp_tstamp = ifp->tstamp;
886 spin_unlock_bh(&ifp->lock);
888 regen_advance = idev->cnf.regen_max_retry *
889 idev->cnf.dad_transmits *
890 idev->nd_parms->retrans_time / HZ;
891 write_unlock(&idev->lock);
893 /* A temporary address is created only if this calculated Preferred
894 * Lifetime is greater than REGEN_ADVANCE time units. In particular,
895 * an implementation must not create a temporary address with a zero
896 * Preferred Lifetime.
898 if (tmp_prefered_lft <= regen_advance) {
899 in6_ifa_put(ifp);
900 in6_dev_put(idev);
901 ret = -1;
902 goto out;
905 addr_flags = IFA_F_TEMPORARY;
906 /* set in addrconf_prefix_rcv() */
907 if (ifp->flags & IFA_F_OPTIMISTIC)
908 addr_flags |= IFA_F_OPTIMISTIC;
910 ift = !max_addresses ||
911 ipv6_count_addresses(idev) < max_addresses ?
912 ipv6_add_addr(idev, &addr, tmp_plen,
913 ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
914 addr_flags) : NULL;
915 if (!ift || IS_ERR(ift)) {
916 in6_ifa_put(ifp);
917 in6_dev_put(idev);
918 printk(KERN_INFO
919 "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
920 tmpaddr = &addr;
921 write_lock(&idev->lock);
922 goto retry;
925 spin_lock_bh(&ift->lock);
926 ift->ifpub = ifp;
927 ift->valid_lft = tmp_valid_lft;
928 ift->prefered_lft = tmp_prefered_lft;
929 ift->cstamp = now;
930 ift->tstamp = tmp_tstamp;
931 spin_unlock_bh(&ift->lock);
933 addrconf_dad_start(ift, 0);
934 in6_ifa_put(ift);
935 in6_dev_put(idev);
936 out:
937 return ret;
939 #endif
942 * Choose an appropriate source address (RFC3484)
944 enum {
945 IPV6_SADDR_RULE_INIT = 0,
946 IPV6_SADDR_RULE_LOCAL,
947 IPV6_SADDR_RULE_SCOPE,
948 IPV6_SADDR_RULE_PREFERRED,
949 #ifdef CONFIG_IPV6_MIP6
950 IPV6_SADDR_RULE_HOA,
951 #endif
952 IPV6_SADDR_RULE_OIF,
953 IPV6_SADDR_RULE_LABEL,
954 #ifdef CONFIG_IPV6_PRIVACY
955 IPV6_SADDR_RULE_PRIVACY,
956 #endif
957 IPV6_SADDR_RULE_ORCHID,
958 IPV6_SADDR_RULE_PREFIX,
959 IPV6_SADDR_RULE_MAX
962 struct ipv6_saddr_score {
963 int rule;
964 int addr_type;
965 struct inet6_ifaddr *ifa;
966 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
967 int scopedist;
968 int matchlen;
971 struct ipv6_saddr_dst {
972 const struct in6_addr *addr;
973 int ifindex;
974 int scope;
975 int label;
976 unsigned int prefs;
979 static inline int ipv6_saddr_preferred(int type)
981 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
982 return 1;
983 return 0;
986 static int ipv6_get_saddr_eval(struct net *net,
987 struct ipv6_saddr_score *score,
988 struct ipv6_saddr_dst *dst,
989 int i)
991 int ret;
993 if (i <= score->rule) {
994 switch (i) {
995 case IPV6_SADDR_RULE_SCOPE:
996 ret = score->scopedist;
997 break;
998 case IPV6_SADDR_RULE_PREFIX:
999 ret = score->matchlen;
1000 break;
1001 default:
1002 ret = !!test_bit(i, score->scorebits);
1004 goto out;
1007 switch (i) {
1008 case IPV6_SADDR_RULE_INIT:
1009 /* Rule 0: remember if hiscore is not ready yet */
1010 ret = !!score->ifa;
1011 break;
1012 case IPV6_SADDR_RULE_LOCAL:
1013 /* Rule 1: Prefer same address */
1014 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1015 break;
1016 case IPV6_SADDR_RULE_SCOPE:
1017 /* Rule 2: Prefer appropriate scope
1019 * ret
1021 * -1 | d 15
1022 * ---+--+-+---> scope
1024 * | d is scope of the destination.
1025 * B-d | \
1026 * | \ <- smaller scope is better if
1027 * B-15 | \ if scope is enough for destinaion.
1028 * | ret = B - scope (-1 <= scope >= d <= 15).
1029 * d-C-1 | /
1030 * |/ <- greater is better
1031 * -C / if scope is not enough for destination.
1032 * /| ret = scope - C (-1 <= d < scope <= 15).
1034 * d - C - 1 < B -15 (for all -1 <= d <= 15).
1035 * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1036 * Assume B = 0 and we get C > 29.
1038 ret = __ipv6_addr_src_scope(score->addr_type);
1039 if (ret >= dst->scope)
1040 ret = -ret;
1041 else
1042 ret -= 128; /* 30 is enough */
1043 score->scopedist = ret;
1044 break;
1045 case IPV6_SADDR_RULE_PREFERRED:
1046 /* Rule 3: Avoid deprecated and optimistic addresses */
1047 ret = ipv6_saddr_preferred(score->addr_type) ||
1048 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1049 break;
1050 #ifdef CONFIG_IPV6_MIP6
1051 case IPV6_SADDR_RULE_HOA:
1053 /* Rule 4: Prefer home address */
1054 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1055 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1056 break;
1058 #endif
1059 case IPV6_SADDR_RULE_OIF:
1060 /* Rule 5: Prefer outgoing interface */
1061 ret = (!dst->ifindex ||
1062 dst->ifindex == score->ifa->idev->dev->ifindex);
1063 break;
1064 case IPV6_SADDR_RULE_LABEL:
1065 /* Rule 6: Prefer matching label */
1066 ret = ipv6_addr_label(net,
1067 &score->ifa->addr, score->addr_type,
1068 score->ifa->idev->dev->ifindex) == dst->label;
1069 break;
1070 #ifdef CONFIG_IPV6_PRIVACY
1071 case IPV6_SADDR_RULE_PRIVACY:
1073 /* Rule 7: Prefer public address
1074 * Note: prefer temporary address if use_tempaddr >= 2
1076 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1077 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1078 score->ifa->idev->cnf.use_tempaddr >= 2;
1079 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1080 break;
1082 #endif
1083 case IPV6_SADDR_RULE_ORCHID:
1084 /* Rule 8-: Prefer ORCHID vs ORCHID or
1085 * non-ORCHID vs non-ORCHID
1087 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1088 ipv6_addr_orchid(dst->addr));
1089 break;
1090 case IPV6_SADDR_RULE_PREFIX:
1091 /* Rule 8: Use longest matching prefix */
1092 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1093 dst->addr);
1094 break;
1095 default:
1096 ret = 0;
1099 if (ret)
1100 __set_bit(i, score->scorebits);
1101 score->rule = i;
1102 out:
1103 return ret;
1106 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1107 const struct in6_addr *daddr, unsigned int prefs,
1108 struct in6_addr *saddr)
1110 struct ipv6_saddr_score scores[2],
1111 *score = &scores[0], *hiscore = &scores[1];
1112 struct ipv6_saddr_dst dst;
1113 struct net_device *dev;
1114 int dst_type;
1116 dst_type = __ipv6_addr_type(daddr);
1117 dst.addr = daddr;
1118 dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1119 dst.scope = __ipv6_addr_src_scope(dst_type);
1120 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1121 dst.prefs = prefs;
1123 hiscore->rule = -1;
1124 hiscore->ifa = NULL;
1126 rcu_read_lock();
1128 for_each_netdev_rcu(net, dev) {
1129 struct inet6_dev *idev;
1131 /* Candidate Source Address (section 4)
1132 * - multicast and link-local destination address,
1133 * the set of candidate source address MUST only
1134 * include addresses assigned to interfaces
1135 * belonging to the same link as the outgoing
1136 * interface.
1137 * (- For site-local destination addresses, the
1138 * set of candidate source addresses MUST only
1139 * include addresses assigned to interfaces
1140 * belonging to the same site as the outgoing
1141 * interface.)
1143 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1144 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1145 dst.ifindex && dev->ifindex != dst.ifindex)
1146 continue;
1148 idev = __in6_dev_get(dev);
1149 if (!idev)
1150 continue;
1152 read_lock_bh(&idev->lock);
1153 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1154 int i;
1157 * - Tentative Address (RFC2462 section 5.4)
1158 * - A tentative address is not considered
1159 * "assigned to an interface" in the traditional
1160 * sense, unless it is also flagged as optimistic.
1161 * - Candidate Source Address (section 4)
1162 * - In any case, anycast addresses, multicast
1163 * addresses, and the unspecified address MUST
1164 * NOT be included in a candidate set.
1166 if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1167 (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1168 continue;
1170 score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1172 if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1173 score->addr_type & IPV6_ADDR_MULTICAST)) {
1174 LIMIT_NETDEBUG(KERN_DEBUG
1175 "ADDRCONF: unspecified / multicast address "
1176 "assigned as unicast address on %s",
1177 dev->name);
1178 continue;
1181 score->rule = -1;
1182 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1184 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1185 int minihiscore, miniscore;
1187 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1188 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1190 if (minihiscore > miniscore) {
1191 if (i == IPV6_SADDR_RULE_SCOPE &&
1192 score->scopedist > 0) {
1194 * special case:
1195 * each remaining entry
1196 * has too small (not enough)
1197 * scope, because ifa entries
1198 * are sorted by their scope
1199 * values.
1201 goto try_nextdev;
1203 break;
1204 } else if (minihiscore < miniscore) {
1205 if (hiscore->ifa)
1206 in6_ifa_put(hiscore->ifa);
1208 in6_ifa_hold(score->ifa);
1210 swap(hiscore, score);
1212 /* restore our iterator */
1213 score->ifa = hiscore->ifa;
1215 break;
1219 try_nextdev:
1220 read_unlock_bh(&idev->lock);
1222 rcu_read_unlock();
1224 if (!hiscore->ifa)
1225 return -EADDRNOTAVAIL;
1227 *saddr = hiscore->ifa->addr;
1228 in6_ifa_put(hiscore->ifa);
1229 return 0;
1231 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1233 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1234 unsigned char banned_flags)
1236 struct inet6_dev *idev;
1237 int err = -EADDRNOTAVAIL;
1239 rcu_read_lock();
1240 idev = __in6_dev_get(dev);
1241 if (idev) {
1242 struct inet6_ifaddr *ifp;
1244 read_lock_bh(&idev->lock);
1245 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1246 if (ifp->scope == IFA_LINK &&
1247 !(ifp->flags & banned_flags)) {
1248 *addr = ifp->addr;
1249 err = 0;
1250 break;
1253 read_unlock_bh(&idev->lock);
1255 rcu_read_unlock();
1256 return err;
1259 static int ipv6_count_addresses(struct inet6_dev *idev)
1261 int cnt = 0;
1262 struct inet6_ifaddr *ifp;
1264 read_lock_bh(&idev->lock);
1265 list_for_each_entry(ifp, &idev->addr_list, if_list)
1266 cnt++;
1267 read_unlock_bh(&idev->lock);
1268 return cnt;
1271 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1272 struct net_device *dev, int strict)
1274 struct inet6_ifaddr *ifp;
1275 struct hlist_node *node;
1276 unsigned int hash = ipv6_addr_hash(addr);
1278 rcu_read_lock_bh();
1279 hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1280 if (!net_eq(dev_net(ifp->idev->dev), net))
1281 continue;
1282 if (ipv6_addr_equal(&ifp->addr, addr) &&
1283 !(ifp->flags&IFA_F_TENTATIVE) &&
1284 (dev == NULL || ifp->idev->dev == dev ||
1285 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1286 rcu_read_unlock_bh();
1287 return 1;
1291 rcu_read_unlock_bh();
1292 return 0;
1294 EXPORT_SYMBOL(ipv6_chk_addr);
1296 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1297 struct net_device *dev)
1299 unsigned int hash = ipv6_addr_hash(addr);
1300 struct inet6_ifaddr *ifp;
1301 struct hlist_node *node;
1303 hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1304 if (!net_eq(dev_net(ifp->idev->dev), net))
1305 continue;
1306 if (ipv6_addr_equal(&ifp->addr, addr)) {
1307 if (dev == NULL || ifp->idev->dev == dev)
1308 return true;
1311 return false;
1314 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1316 struct inet6_dev *idev;
1317 struct inet6_ifaddr *ifa;
1318 int onlink;
1320 onlink = 0;
1321 rcu_read_lock();
1322 idev = __in6_dev_get(dev);
1323 if (idev) {
1324 read_lock_bh(&idev->lock);
1325 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1326 onlink = ipv6_prefix_equal(addr, &ifa->addr,
1327 ifa->prefix_len);
1328 if (onlink)
1329 break;
1331 read_unlock_bh(&idev->lock);
1333 rcu_read_unlock();
1334 return onlink;
1337 EXPORT_SYMBOL(ipv6_chk_prefix);
1339 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1340 struct net_device *dev, int strict)
1342 struct inet6_ifaddr *ifp, *result = NULL;
1343 unsigned int hash = ipv6_addr_hash(addr);
1344 struct hlist_node *node;
1346 rcu_read_lock_bh();
1347 hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1348 if (!net_eq(dev_net(ifp->idev->dev), net))
1349 continue;
1350 if (ipv6_addr_equal(&ifp->addr, addr)) {
1351 if (dev == NULL || ifp->idev->dev == dev ||
1352 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1353 result = ifp;
1354 in6_ifa_hold(ifp);
1355 break;
1359 rcu_read_unlock_bh();
1361 return result;
1364 /* Gets referenced address, destroys ifaddr */
1366 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1368 if (ifp->flags&IFA_F_PERMANENT) {
1369 spin_lock_bh(&ifp->lock);
1370 addrconf_del_timer(ifp);
1371 ifp->flags |= IFA_F_TENTATIVE;
1372 if (dad_failed)
1373 ifp->flags |= IFA_F_DADFAILED;
1374 spin_unlock_bh(&ifp->lock);
1375 if (dad_failed)
1376 ipv6_ifa_notify(0, ifp);
1377 in6_ifa_put(ifp);
1378 #ifdef CONFIG_IPV6_PRIVACY
1379 } else if (ifp->flags&IFA_F_TEMPORARY) {
1380 struct inet6_ifaddr *ifpub;
1381 spin_lock_bh(&ifp->lock);
1382 ifpub = ifp->ifpub;
1383 if (ifpub) {
1384 in6_ifa_hold(ifpub);
1385 spin_unlock_bh(&ifp->lock);
1386 ipv6_create_tempaddr(ifpub, ifp);
1387 in6_ifa_put(ifpub);
1388 } else {
1389 spin_unlock_bh(&ifp->lock);
1391 ipv6_del_addr(ifp);
1392 #endif
1393 } else
1394 ipv6_del_addr(ifp);
1397 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1399 int err = -ENOENT;
1401 spin_lock(&ifp->state_lock);
1402 if (ifp->state == INET6_IFADDR_STATE_DAD) {
1403 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1404 err = 0;
1406 spin_unlock(&ifp->state_lock);
1408 return err;
1411 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1413 struct inet6_dev *idev = ifp->idev;
1415 if (addrconf_dad_end(ifp)) {
1416 in6_ifa_put(ifp);
1417 return;
1420 if (net_ratelimit())
1421 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1422 ifp->idev->dev->name, &ifp->addr);
1424 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1425 struct in6_addr addr;
1427 addr.s6_addr32[0] = htonl(0xfe800000);
1428 addr.s6_addr32[1] = 0;
1430 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1431 ipv6_addr_equal(&ifp->addr, &addr)) {
1432 /* DAD failed for link-local based on MAC address */
1433 idev->cnf.disable_ipv6 = 1;
1435 printk(KERN_INFO "%s: IPv6 being disabled!\n",
1436 ifp->idev->dev->name);
1440 addrconf_dad_stop(ifp, 1);
1443 /* Join to solicited addr multicast group. */
1445 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1447 struct in6_addr maddr;
1449 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1450 return;
1452 addrconf_addr_solict_mult(addr, &maddr);
1453 ipv6_dev_mc_inc(dev, &maddr);
1456 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1458 struct in6_addr maddr;
1460 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1461 return;
1463 addrconf_addr_solict_mult(addr, &maddr);
1464 __ipv6_dev_mc_dec(idev, &maddr);
1467 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1469 struct in6_addr addr;
1470 if (ifp->prefix_len == 127) /* RFC 6164 */
1471 return;
1472 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1473 if (ipv6_addr_any(&addr))
1474 return;
1475 ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1478 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1480 struct in6_addr addr;
1481 if (ifp->prefix_len == 127) /* RFC 6164 */
1482 return;
1483 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1484 if (ipv6_addr_any(&addr))
1485 return;
1486 __ipv6_dev_ac_dec(ifp->idev, &addr);
1489 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1491 if (dev->addr_len != ETH_ALEN)
1492 return -1;
1493 memcpy(eui, dev->dev_addr, 3);
1494 memcpy(eui + 5, dev->dev_addr + 3, 3);
1497 * The zSeries OSA network cards can be shared among various
1498 * OS instances, but the OSA cards have only one MAC address.
1499 * This leads to duplicate address conflicts in conjunction
1500 * with IPv6 if more than one instance uses the same card.
1502 * The driver for these cards can deliver a unique 16-bit
1503 * identifier for each instance sharing the same card. It is
1504 * placed instead of 0xFFFE in the interface identifier. The
1505 * "u" bit of the interface identifier is not inverted in this
1506 * case. Hence the resulting interface identifier has local
1507 * scope according to RFC2373.
1509 if (dev->dev_id) {
1510 eui[3] = (dev->dev_id >> 8) & 0xFF;
1511 eui[4] = dev->dev_id & 0xFF;
1512 } else {
1513 eui[3] = 0xFF;
1514 eui[4] = 0xFE;
1515 eui[0] ^= 2;
1517 return 0;
1520 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1522 /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1523 if (dev->addr_len != ARCNET_ALEN)
1524 return -1;
1525 memset(eui, 0, 7);
1526 eui[7] = *(u8*)dev->dev_addr;
1527 return 0;
1530 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1532 if (dev->addr_len != INFINIBAND_ALEN)
1533 return -1;
1534 memcpy(eui, dev->dev_addr + 12, 8);
1535 eui[0] |= 2;
1536 return 0;
1539 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1541 if (addr == 0)
1542 return -1;
1543 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1544 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1545 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1546 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1547 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1548 ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1549 eui[1] = 0;
1550 eui[2] = 0x5E;
1551 eui[3] = 0xFE;
1552 memcpy(eui + 4, &addr, 4);
1553 return 0;
1556 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1558 if (dev->priv_flags & IFF_ISATAP)
1559 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1560 return -1;
1563 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1565 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1568 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1570 switch (dev->type) {
1571 case ARPHRD_ETHER:
1572 case ARPHRD_FDDI:
1573 case ARPHRD_IEEE802_TR:
1574 return addrconf_ifid_eui48(eui, dev);
1575 case ARPHRD_ARCNET:
1576 return addrconf_ifid_arcnet(eui, dev);
1577 case ARPHRD_INFINIBAND:
1578 return addrconf_ifid_infiniband(eui, dev);
1579 case ARPHRD_SIT:
1580 return addrconf_ifid_sit(eui, dev);
1581 case ARPHRD_IPGRE:
1582 return addrconf_ifid_gre(eui, dev);
1584 return -1;
1587 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1589 int err = -1;
1590 struct inet6_ifaddr *ifp;
1592 read_lock_bh(&idev->lock);
1593 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1594 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1595 memcpy(eui, ifp->addr.s6_addr+8, 8);
1596 err = 0;
1597 break;
1600 read_unlock_bh(&idev->lock);
1601 return err;
1604 #ifdef CONFIG_IPV6_PRIVACY
1605 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1606 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1608 regen:
1609 get_random_bytes(idev->rndid, sizeof(idev->rndid));
1610 idev->rndid[0] &= ~0x02;
1613 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1614 * check if generated address is not inappropriate
1616 * - Reserved subnet anycast (RFC 2526)
1617 * 11111101 11....11 1xxxxxxx
1618 * - ISATAP (RFC4214) 6.1
1619 * 00-00-5E-FE-xx-xx-xx-xx
1620 * - value 0
1621 * - XXX: already assigned to an address on the device
1623 if (idev->rndid[0] == 0xfd &&
1624 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1625 (idev->rndid[7]&0x80))
1626 goto regen;
1627 if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1628 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1629 goto regen;
1630 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1631 goto regen;
1634 return 0;
1637 static void ipv6_regen_rndid(unsigned long data)
1639 struct inet6_dev *idev = (struct inet6_dev *) data;
1640 unsigned long expires;
1642 rcu_read_lock_bh();
1643 write_lock_bh(&idev->lock);
1645 if (idev->dead)
1646 goto out;
1648 if (__ipv6_regen_rndid(idev) < 0)
1649 goto out;
1651 expires = jiffies +
1652 idev->cnf.temp_prefered_lft * HZ -
1653 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1654 idev->cnf.max_desync_factor * HZ;
1655 if (time_before(expires, jiffies)) {
1656 printk(KERN_WARNING
1657 "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1658 idev->dev->name);
1659 goto out;
1662 if (!mod_timer(&idev->regen_timer, expires))
1663 in6_dev_hold(idev);
1665 out:
1666 write_unlock_bh(&idev->lock);
1667 rcu_read_unlock_bh();
1668 in6_dev_put(idev);
1671 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1672 int ret = 0;
1674 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1675 ret = __ipv6_regen_rndid(idev);
1676 return ret;
1678 #endif
1681 * Add prefix route.
1684 static void
1685 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1686 unsigned long expires, u32 flags)
1688 struct fib6_config cfg = {
1689 .fc_table = RT6_TABLE_PREFIX,
1690 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1691 .fc_ifindex = dev->ifindex,
1692 .fc_expires = expires,
1693 .fc_dst_len = plen,
1694 .fc_flags = RTF_UP | flags,
1695 .fc_nlinfo.nl_net = dev_net(dev),
1696 .fc_protocol = RTPROT_KERNEL,
1699 cfg.fc_dst = *pfx;
1701 /* Prevent useless cloning on PtP SIT.
1702 This thing is done here expecting that the whole
1703 class of non-broadcast devices need not cloning.
1705 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1706 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1707 cfg.fc_flags |= RTF_NONEXTHOP;
1708 #endif
1710 ip6_route_add(&cfg);
1714 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1715 int plen,
1716 const struct net_device *dev,
1717 u32 flags, u32 noflags)
1719 struct fib6_node *fn;
1720 struct rt6_info *rt = NULL;
1721 struct fib6_table *table;
1723 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1724 if (table == NULL)
1725 return NULL;
1727 write_lock_bh(&table->tb6_lock);
1728 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1729 if (!fn)
1730 goto out;
1731 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1732 if (rt->dst.dev->ifindex != dev->ifindex)
1733 continue;
1734 if ((rt->rt6i_flags & flags) != flags)
1735 continue;
1736 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1737 continue;
1738 dst_hold(&rt->dst);
1739 break;
1741 out:
1742 write_unlock_bh(&table->tb6_lock);
1743 return rt;
1747 /* Create "default" multicast route to the interface */
1749 static void addrconf_add_mroute(struct net_device *dev)
1751 struct fib6_config cfg = {
1752 .fc_table = RT6_TABLE_LOCAL,
1753 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1754 .fc_ifindex = dev->ifindex,
1755 .fc_dst_len = 8,
1756 .fc_flags = RTF_UP,
1757 .fc_nlinfo.nl_net = dev_net(dev),
1760 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1762 ip6_route_add(&cfg);
1765 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1766 static void sit_route_add(struct net_device *dev)
1768 struct fib6_config cfg = {
1769 .fc_table = RT6_TABLE_MAIN,
1770 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1771 .fc_ifindex = dev->ifindex,
1772 .fc_dst_len = 96,
1773 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1774 .fc_nlinfo.nl_net = dev_net(dev),
1777 /* prefix length - 96 bits "::d.d.d.d" */
1778 ip6_route_add(&cfg);
1780 #endif
1782 static void addrconf_add_lroute(struct net_device *dev)
1784 struct in6_addr addr;
1786 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
1787 addrconf_prefix_route(&addr, 64, dev, 0, 0);
1790 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1792 struct inet6_dev *idev;
1794 ASSERT_RTNL();
1796 idev = ipv6_find_idev(dev);
1797 if (!idev)
1798 return ERR_PTR(-ENOBUFS);
1800 if (idev->cnf.disable_ipv6)
1801 return ERR_PTR(-EACCES);
1803 /* Add default multicast route */
1804 if (!(dev->flags & IFF_LOOPBACK))
1805 addrconf_add_mroute(dev);
1807 /* Add link local route */
1808 addrconf_add_lroute(dev);
1809 return idev;
1812 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao)
1814 struct prefix_info *pinfo;
1815 __u32 valid_lft;
1816 __u32 prefered_lft;
1817 int addr_type;
1818 struct inet6_dev *in6_dev;
1819 struct net *net = dev_net(dev);
1821 pinfo = (struct prefix_info *) opt;
1823 if (len < sizeof(struct prefix_info)) {
1824 ADBG(("addrconf: prefix option too short\n"));
1825 return;
1829 * Validation checks ([ADDRCONF], page 19)
1832 addr_type = ipv6_addr_type(&pinfo->prefix);
1834 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1835 return;
1837 valid_lft = ntohl(pinfo->valid);
1838 prefered_lft = ntohl(pinfo->prefered);
1840 if (prefered_lft > valid_lft) {
1841 if (net_ratelimit())
1842 printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1843 return;
1846 in6_dev = in6_dev_get(dev);
1848 if (in6_dev == NULL) {
1849 if (net_ratelimit())
1850 printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1851 return;
1855 * Two things going on here:
1856 * 1) Add routes for on-link prefixes
1857 * 2) Configure prefixes with the auto flag set
1860 if (pinfo->onlink) {
1861 struct rt6_info *rt;
1862 unsigned long rt_expires;
1864 /* Avoid arithmetic overflow. Really, we could
1865 * save rt_expires in seconds, likely valid_lft,
1866 * but it would require division in fib gc, that it
1867 * not good.
1869 if (HZ > USER_HZ)
1870 rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1871 else
1872 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1874 if (addrconf_finite_timeout(rt_expires))
1875 rt_expires *= HZ;
1877 rt = addrconf_get_prefix_route(&pinfo->prefix,
1878 pinfo->prefix_len,
1879 dev,
1880 RTF_ADDRCONF | RTF_PREFIX_RT,
1881 RTF_GATEWAY | RTF_DEFAULT);
1883 if (rt) {
1884 /* Autoconf prefix route */
1885 if (valid_lft == 0) {
1886 ip6_del_rt(rt);
1887 rt = NULL;
1888 } else if (addrconf_finite_timeout(rt_expires)) {
1889 /* not infinity */
1890 rt->dst.expires = jiffies + rt_expires;
1891 rt->rt6i_flags |= RTF_EXPIRES;
1892 } else {
1893 rt->rt6i_flags &= ~RTF_EXPIRES;
1894 rt->dst.expires = 0;
1896 } else if (valid_lft) {
1897 clock_t expires = 0;
1898 int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1899 if (addrconf_finite_timeout(rt_expires)) {
1900 /* not infinity */
1901 flags |= RTF_EXPIRES;
1902 expires = jiffies_to_clock_t(rt_expires);
1904 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1905 dev, expires, flags);
1907 if (rt)
1908 dst_release(&rt->dst);
1911 /* Try to figure out our local address for this prefix */
1913 if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1914 struct inet6_ifaddr * ifp;
1915 struct in6_addr addr;
1916 int create = 0, update_lft = 0;
1918 if (pinfo->prefix_len == 64) {
1919 memcpy(&addr, &pinfo->prefix, 8);
1920 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1921 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1922 in6_dev_put(in6_dev);
1923 return;
1925 goto ok;
1927 if (net_ratelimit())
1928 printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1929 pinfo->prefix_len);
1930 in6_dev_put(in6_dev);
1931 return;
1935 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1937 if (ifp == NULL && valid_lft) {
1938 int max_addresses = in6_dev->cnf.max_addresses;
1939 u32 addr_flags = 0;
1941 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1942 if (in6_dev->cnf.optimistic_dad &&
1943 !net->ipv6.devconf_all->forwarding && sllao)
1944 addr_flags = IFA_F_OPTIMISTIC;
1945 #endif
1947 /* Do not allow to create too much of autoconfigured
1948 * addresses; this would be too easy way to crash kernel.
1950 if (!max_addresses ||
1951 ipv6_count_addresses(in6_dev) < max_addresses)
1952 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1953 addr_type&IPV6_ADDR_SCOPE_MASK,
1954 addr_flags);
1956 if (!ifp || IS_ERR(ifp)) {
1957 in6_dev_put(in6_dev);
1958 return;
1961 update_lft = create = 1;
1962 ifp->cstamp = jiffies;
1963 addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1966 if (ifp) {
1967 int flags;
1968 unsigned long now;
1969 #ifdef CONFIG_IPV6_PRIVACY
1970 struct inet6_ifaddr *ift;
1971 #endif
1972 u32 stored_lft;
1974 /* update lifetime (RFC2462 5.5.3 e) */
1975 spin_lock(&ifp->lock);
1976 now = jiffies;
1977 if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1978 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1979 else
1980 stored_lft = 0;
1981 if (!update_lft && stored_lft) {
1982 if (valid_lft > MIN_VALID_LIFETIME ||
1983 valid_lft > stored_lft)
1984 update_lft = 1;
1985 else if (stored_lft <= MIN_VALID_LIFETIME) {
1986 /* valid_lft <= stored_lft is always true */
1988 * RFC 4862 Section 5.5.3e:
1989 * "Note that the preferred lifetime of
1990 * the corresponding address is always
1991 * reset to the Preferred Lifetime in
1992 * the received Prefix Information
1993 * option, regardless of whether the
1994 * valid lifetime is also reset or
1995 * ignored."
1997 * So if the preferred lifetime in
1998 * this advertisement is different
1999 * than what we have stored, but the
2000 * valid lifetime is invalid, just
2001 * reset prefered_lft.
2003 * We must set the valid lifetime
2004 * to the stored lifetime since we'll
2005 * be updating the timestamp below,
2006 * else we'll set it back to the
2007 * minimum.
2009 if (prefered_lft != ifp->prefered_lft) {
2010 valid_lft = stored_lft;
2011 update_lft = 1;
2013 } else {
2014 valid_lft = MIN_VALID_LIFETIME;
2015 if (valid_lft < prefered_lft)
2016 prefered_lft = valid_lft;
2017 update_lft = 1;
2021 if (update_lft) {
2022 ifp->valid_lft = valid_lft;
2023 ifp->prefered_lft = prefered_lft;
2024 ifp->tstamp = now;
2025 flags = ifp->flags;
2026 ifp->flags &= ~IFA_F_DEPRECATED;
2027 spin_unlock(&ifp->lock);
2029 if (!(flags&IFA_F_TENTATIVE))
2030 ipv6_ifa_notify(0, ifp);
2031 } else
2032 spin_unlock(&ifp->lock);
2034 #ifdef CONFIG_IPV6_PRIVACY
2035 read_lock_bh(&in6_dev->lock);
2036 /* update all temporary addresses in the list */
2037 list_for_each_entry(ift, &in6_dev->tempaddr_list,
2038 tmp_list) {
2039 int age, max_valid, max_prefered;
2041 if (ifp != ift->ifpub)
2042 continue;
2045 * RFC 4941 section 3.3:
2046 * If a received option will extend the lifetime
2047 * of a public address, the lifetimes of
2048 * temporary addresses should be extended,
2049 * subject to the overall constraint that no
2050 * temporary addresses should ever remain
2051 * "valid" or "preferred" for a time longer than
2052 * (TEMP_VALID_LIFETIME) or
2053 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2054 * respectively.
2056 age = (now - ift->cstamp) / HZ;
2057 max_valid = in6_dev->cnf.temp_valid_lft - age;
2058 if (max_valid < 0)
2059 max_valid = 0;
2061 max_prefered = in6_dev->cnf.temp_prefered_lft -
2062 in6_dev->cnf.max_desync_factor -
2063 age;
2064 if (max_prefered < 0)
2065 max_prefered = 0;
2067 if (valid_lft > max_valid)
2068 valid_lft = max_valid;
2070 if (prefered_lft > max_prefered)
2071 prefered_lft = max_prefered;
2073 spin_lock(&ift->lock);
2074 flags = ift->flags;
2075 ift->valid_lft = valid_lft;
2076 ift->prefered_lft = prefered_lft;
2077 ift->tstamp = now;
2078 if (prefered_lft > 0)
2079 ift->flags &= ~IFA_F_DEPRECATED;
2081 spin_unlock(&ift->lock);
2082 if (!(flags&IFA_F_TENTATIVE))
2083 ipv6_ifa_notify(0, ift);
2086 if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2088 * When a new public address is created as
2089 * described in [ADDRCONF], also create a new
2090 * temporary address. Also create a temporary
2091 * address if it's enabled but no temporary
2092 * address currently exists.
2094 read_unlock_bh(&in6_dev->lock);
2095 ipv6_create_tempaddr(ifp, NULL);
2096 } else {
2097 read_unlock_bh(&in6_dev->lock);
2099 #endif
2100 in6_ifa_put(ifp);
2101 addrconf_verify(0);
2104 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2105 in6_dev_put(in6_dev);
2109 * Set destination address.
2110 * Special case for SIT interfaces where we create a new "virtual"
2111 * device.
2113 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2115 struct in6_ifreq ireq;
2116 struct net_device *dev;
2117 int err = -EINVAL;
2119 rtnl_lock();
2121 err = -EFAULT;
2122 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2123 goto err_exit;
2125 dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2127 err = -ENODEV;
2128 if (dev == NULL)
2129 goto err_exit;
2131 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2132 if (dev->type == ARPHRD_SIT) {
2133 const struct net_device_ops *ops = dev->netdev_ops;
2134 struct ifreq ifr;
2135 struct ip_tunnel_parm p;
2137 err = -EADDRNOTAVAIL;
2138 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2139 goto err_exit;
2141 memset(&p, 0, sizeof(p));
2142 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2143 p.iph.saddr = 0;
2144 p.iph.version = 4;
2145 p.iph.ihl = 5;
2146 p.iph.protocol = IPPROTO_IPV6;
2147 p.iph.ttl = 64;
2148 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2150 if (ops->ndo_do_ioctl) {
2151 mm_segment_t oldfs = get_fs();
2153 set_fs(KERNEL_DS);
2154 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2155 set_fs(oldfs);
2156 } else
2157 err = -EOPNOTSUPP;
2159 if (err == 0) {
2160 err = -ENOBUFS;
2161 dev = __dev_get_by_name(net, p.name);
2162 if (!dev)
2163 goto err_exit;
2164 err = dev_open(dev);
2167 #endif
2169 err_exit:
2170 rtnl_unlock();
2171 return err;
2175 * Manual configuration of address on an interface
2177 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2178 unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2179 __u32 valid_lft)
2181 struct inet6_ifaddr *ifp;
2182 struct inet6_dev *idev;
2183 struct net_device *dev;
2184 int scope;
2185 u32 flags;
2186 clock_t expires;
2187 unsigned long timeout;
2189 ASSERT_RTNL();
2191 if (plen > 128)
2192 return -EINVAL;
2194 /* check the lifetime */
2195 if (!valid_lft || prefered_lft > valid_lft)
2196 return -EINVAL;
2198 dev = __dev_get_by_index(net, ifindex);
2199 if (!dev)
2200 return -ENODEV;
2202 idev = addrconf_add_dev(dev);
2203 if (IS_ERR(idev))
2204 return PTR_ERR(idev);
2206 scope = ipv6_addr_scope(pfx);
2208 timeout = addrconf_timeout_fixup(valid_lft, HZ);
2209 if (addrconf_finite_timeout(timeout)) {
2210 expires = jiffies_to_clock_t(timeout * HZ);
2211 valid_lft = timeout;
2212 flags = RTF_EXPIRES;
2213 } else {
2214 expires = 0;
2215 flags = 0;
2216 ifa_flags |= IFA_F_PERMANENT;
2219 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2220 if (addrconf_finite_timeout(timeout)) {
2221 if (timeout == 0)
2222 ifa_flags |= IFA_F_DEPRECATED;
2223 prefered_lft = timeout;
2226 ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2228 if (!IS_ERR(ifp)) {
2229 spin_lock_bh(&ifp->lock);
2230 ifp->valid_lft = valid_lft;
2231 ifp->prefered_lft = prefered_lft;
2232 ifp->tstamp = jiffies;
2233 spin_unlock_bh(&ifp->lock);
2235 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2236 expires, flags);
2238 * Note that section 3.1 of RFC 4429 indicates
2239 * that the Optimistic flag should not be set for
2240 * manually configured addresses
2242 addrconf_dad_start(ifp, 0);
2243 in6_ifa_put(ifp);
2244 addrconf_verify(0);
2245 return 0;
2248 return PTR_ERR(ifp);
2251 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2252 unsigned int plen)
2254 struct inet6_ifaddr *ifp;
2255 struct inet6_dev *idev;
2256 struct net_device *dev;
2258 if (plen > 128)
2259 return -EINVAL;
2261 dev = __dev_get_by_index(net, ifindex);
2262 if (!dev)
2263 return -ENODEV;
2265 if ((idev = __in6_dev_get(dev)) == NULL)
2266 return -ENXIO;
2268 read_lock_bh(&idev->lock);
2269 list_for_each_entry(ifp, &idev->addr_list, if_list) {
2270 if (ifp->prefix_len == plen &&
2271 ipv6_addr_equal(pfx, &ifp->addr)) {
2272 in6_ifa_hold(ifp);
2273 read_unlock_bh(&idev->lock);
2275 ipv6_del_addr(ifp);
2277 /* If the last address is deleted administratively,
2278 disable IPv6 on this interface.
2280 if (list_empty(&idev->addr_list))
2281 addrconf_ifdown(idev->dev, 1);
2282 return 0;
2285 read_unlock_bh(&idev->lock);
2286 return -EADDRNOTAVAIL;
2290 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2292 struct in6_ifreq ireq;
2293 int err;
2295 if (!capable(CAP_NET_ADMIN))
2296 return -EPERM;
2298 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2299 return -EFAULT;
2301 rtnl_lock();
2302 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2303 ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2304 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2305 rtnl_unlock();
2306 return err;
2309 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2311 struct in6_ifreq ireq;
2312 int err;
2314 if (!capable(CAP_NET_ADMIN))
2315 return -EPERM;
2317 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2318 return -EFAULT;
2320 rtnl_lock();
2321 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2322 ireq.ifr6_prefixlen);
2323 rtnl_unlock();
2324 return err;
2327 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2328 int plen, int scope)
2330 struct inet6_ifaddr *ifp;
2332 ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2333 if (!IS_ERR(ifp)) {
2334 spin_lock_bh(&ifp->lock);
2335 ifp->flags &= ~IFA_F_TENTATIVE;
2336 spin_unlock_bh(&ifp->lock);
2337 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2338 in6_ifa_put(ifp);
2342 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2343 static void sit_add_v4_addrs(struct inet6_dev *idev)
2345 struct in6_addr addr;
2346 struct net_device *dev;
2347 struct net *net = dev_net(idev->dev);
2348 int scope;
2350 ASSERT_RTNL();
2352 memset(&addr, 0, sizeof(struct in6_addr));
2353 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2355 if (idev->dev->flags&IFF_POINTOPOINT) {
2356 addr.s6_addr32[0] = htonl(0xfe800000);
2357 scope = IFA_LINK;
2358 } else {
2359 scope = IPV6_ADDR_COMPATv4;
2362 if (addr.s6_addr32[3]) {
2363 add_addr(idev, &addr, 128, scope);
2364 return;
2367 for_each_netdev(net, dev) {
2368 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2369 if (in_dev && (dev->flags & IFF_UP)) {
2370 struct in_ifaddr * ifa;
2372 int flag = scope;
2374 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2375 int plen;
2377 addr.s6_addr32[3] = ifa->ifa_local;
2379 if (ifa->ifa_scope == RT_SCOPE_LINK)
2380 continue;
2381 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2382 if (idev->dev->flags&IFF_POINTOPOINT)
2383 continue;
2384 flag |= IFA_HOST;
2386 if (idev->dev->flags&IFF_POINTOPOINT)
2387 plen = 64;
2388 else
2389 plen = 96;
2391 add_addr(idev, &addr, plen, flag);
2396 #endif
2398 static void init_loopback(struct net_device *dev)
2400 struct inet6_dev *idev;
2402 /* ::1 */
2404 ASSERT_RTNL();
2406 if ((idev = ipv6_find_idev(dev)) == NULL) {
2407 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2408 return;
2411 add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2414 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2416 struct inet6_ifaddr * ifp;
2417 u32 addr_flags = IFA_F_PERMANENT;
2419 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2420 if (idev->cnf.optimistic_dad &&
2421 !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2422 addr_flags |= IFA_F_OPTIMISTIC;
2423 #endif
2426 ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2427 if (!IS_ERR(ifp)) {
2428 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2429 addrconf_dad_start(ifp, 0);
2430 in6_ifa_put(ifp);
2434 static void addrconf_dev_config(struct net_device *dev)
2436 struct in6_addr addr;
2437 struct inet6_dev * idev;
2439 ASSERT_RTNL();
2441 if ((dev->type != ARPHRD_ETHER) &&
2442 (dev->type != ARPHRD_FDDI) &&
2443 (dev->type != ARPHRD_IEEE802_TR) &&
2444 (dev->type != ARPHRD_ARCNET) &&
2445 (dev->type != ARPHRD_INFINIBAND)) {
2446 /* Alas, we support only Ethernet autoconfiguration. */
2447 return;
2450 idev = addrconf_add_dev(dev);
2451 if (IS_ERR(idev))
2452 return;
2454 memset(&addr, 0, sizeof(struct in6_addr));
2455 addr.s6_addr32[0] = htonl(0xFE800000);
2457 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2458 addrconf_add_linklocal(idev, &addr);
2461 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2462 static void addrconf_sit_config(struct net_device *dev)
2464 struct inet6_dev *idev;
2466 ASSERT_RTNL();
2469 * Configure the tunnel with one of our IPv4
2470 * addresses... we should configure all of
2471 * our v4 addrs in the tunnel
2474 if ((idev = ipv6_find_idev(dev)) == NULL) {
2475 printk(KERN_DEBUG "init sit: add_dev failed\n");
2476 return;
2479 if (dev->priv_flags & IFF_ISATAP) {
2480 struct in6_addr addr;
2482 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2483 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2484 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2485 addrconf_add_linklocal(idev, &addr);
2486 return;
2489 sit_add_v4_addrs(idev);
2491 if (dev->flags&IFF_POINTOPOINT) {
2492 addrconf_add_mroute(dev);
2493 addrconf_add_lroute(dev);
2494 } else
2495 sit_route_add(dev);
2497 #endif
2499 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2500 static void addrconf_gre_config(struct net_device *dev)
2502 struct inet6_dev *idev;
2503 struct in6_addr addr;
2505 pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2507 ASSERT_RTNL();
2509 if ((idev = ipv6_find_idev(dev)) == NULL) {
2510 printk(KERN_DEBUG "init gre: add_dev failed\n");
2511 return;
2514 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0);
2515 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2517 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2518 addrconf_add_linklocal(idev, &addr);
2520 #endif
2522 static inline int
2523 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2525 struct in6_addr lladdr;
2527 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2528 addrconf_add_linklocal(idev, &lladdr);
2529 return 0;
2531 return -1;
2534 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2536 struct net_device *link_dev;
2537 struct net *net = dev_net(idev->dev);
2539 /* first try to inherit the link-local address from the link device */
2540 if (idev->dev->iflink &&
2541 (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2542 if (!ipv6_inherit_linklocal(idev, link_dev))
2543 return;
2545 /* then try to inherit it from any device */
2546 for_each_netdev(net, link_dev) {
2547 if (!ipv6_inherit_linklocal(idev, link_dev))
2548 return;
2550 printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2554 * Autoconfigure tunnel with a link-local address so routing protocols,
2555 * DHCPv6, MLD etc. can be run over the virtual link
2558 static void addrconf_ip6_tnl_config(struct net_device *dev)
2560 struct inet6_dev *idev;
2562 ASSERT_RTNL();
2564 idev = addrconf_add_dev(dev);
2565 if (IS_ERR(idev)) {
2566 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2567 return;
2569 ip6_tnl_add_linklocal(idev);
2572 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2573 void * data)
2575 struct net_device *dev = (struct net_device *) data;
2576 struct inet6_dev *idev = __in6_dev_get(dev);
2577 int run_pending = 0;
2578 int err;
2580 switch (event) {
2581 case NETDEV_REGISTER:
2582 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2583 idev = ipv6_add_dev(dev);
2584 if (!idev)
2585 return notifier_from_errno(-ENOMEM);
2587 break;
2589 case NETDEV_UP:
2590 case NETDEV_CHANGE:
2591 if (dev->flags & IFF_SLAVE)
2592 break;
2594 if (event == NETDEV_UP) {
2595 if (!addrconf_qdisc_ok(dev)) {
2596 /* device is not ready yet. */
2597 printk(KERN_INFO
2598 "ADDRCONF(NETDEV_UP): %s: "
2599 "link is not ready\n",
2600 dev->name);
2601 break;
2604 if (!idev && dev->mtu >= IPV6_MIN_MTU)
2605 idev = ipv6_add_dev(dev);
2607 if (idev) {
2608 idev->if_flags |= IF_READY;
2609 run_pending = 1;
2611 } else {
2612 if (!addrconf_qdisc_ok(dev)) {
2613 /* device is still not ready. */
2614 break;
2617 if (idev) {
2618 if (idev->if_flags & IF_READY)
2619 /* device is already configured. */
2620 break;
2621 idev->if_flags |= IF_READY;
2624 printk(KERN_INFO
2625 "ADDRCONF(NETDEV_CHANGE): %s: "
2626 "link becomes ready\n",
2627 dev->name);
2629 run_pending = 1;
2632 switch (dev->type) {
2633 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2634 case ARPHRD_SIT:
2635 addrconf_sit_config(dev);
2636 break;
2637 #endif
2638 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2639 case ARPHRD_IPGRE:
2640 addrconf_gre_config(dev);
2641 break;
2642 #endif
2643 case ARPHRD_TUNNEL6:
2644 addrconf_ip6_tnl_config(dev);
2645 break;
2646 case ARPHRD_LOOPBACK:
2647 init_loopback(dev);
2648 break;
2650 default:
2651 addrconf_dev_config(dev);
2652 break;
2655 if (idev) {
2656 if (run_pending)
2657 addrconf_dad_run(idev);
2660 * If the MTU changed during the interface down,
2661 * when the interface up, the changed MTU must be
2662 * reflected in the idev as well as routers.
2664 if (idev->cnf.mtu6 != dev->mtu &&
2665 dev->mtu >= IPV6_MIN_MTU) {
2666 rt6_mtu_change(dev, dev->mtu);
2667 idev->cnf.mtu6 = dev->mtu;
2669 idev->tstamp = jiffies;
2670 inet6_ifinfo_notify(RTM_NEWLINK, idev);
2673 * If the changed mtu during down is lower than
2674 * IPV6_MIN_MTU stop IPv6 on this interface.
2676 if (dev->mtu < IPV6_MIN_MTU)
2677 addrconf_ifdown(dev, 1);
2679 break;
2681 case NETDEV_CHANGEMTU:
2682 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2683 rt6_mtu_change(dev, dev->mtu);
2684 idev->cnf.mtu6 = dev->mtu;
2685 break;
2688 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2689 idev = ipv6_add_dev(dev);
2690 if (idev)
2691 break;
2695 * MTU falled under IPV6_MIN_MTU.
2696 * Stop IPv6 on this interface.
2699 case NETDEV_DOWN:
2700 case NETDEV_UNREGISTER:
2702 * Remove all addresses from this interface.
2704 addrconf_ifdown(dev, event != NETDEV_DOWN);
2705 break;
2707 case NETDEV_CHANGENAME:
2708 if (idev) {
2709 snmp6_unregister_dev(idev);
2710 addrconf_sysctl_unregister(idev);
2711 addrconf_sysctl_register(idev);
2712 err = snmp6_register_dev(idev);
2713 if (err)
2714 return notifier_from_errno(err);
2716 break;
2718 case NETDEV_PRE_TYPE_CHANGE:
2719 case NETDEV_POST_TYPE_CHANGE:
2720 addrconf_type_change(dev, event);
2721 break;
2724 return NOTIFY_OK;
2728 * addrconf module should be notified of a device going up
2730 static struct notifier_block ipv6_dev_notf = {
2731 .notifier_call = addrconf_notify,
2734 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2736 struct inet6_dev *idev;
2737 ASSERT_RTNL();
2739 idev = __in6_dev_get(dev);
2741 if (event == NETDEV_POST_TYPE_CHANGE)
2742 ipv6_mc_remap(idev);
2743 else if (event == NETDEV_PRE_TYPE_CHANGE)
2744 ipv6_mc_unmap(idev);
2747 static int addrconf_ifdown(struct net_device *dev, int how)
2749 struct net *net = dev_net(dev);
2750 struct inet6_dev *idev;
2751 struct inet6_ifaddr *ifa;
2752 int state, i;
2754 ASSERT_RTNL();
2756 rt6_ifdown(net, dev);
2757 neigh_ifdown(&nd_tbl, dev);
2759 idev = __in6_dev_get(dev);
2760 if (idev == NULL)
2761 return -ENODEV;
2764 * Step 1: remove reference to ipv6 device from parent device.
2765 * Do not dev_put!
2767 if (how) {
2768 idev->dead = 1;
2770 /* protected by rtnl_lock */
2771 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2773 /* Step 1.5: remove snmp6 entry */
2774 snmp6_unregister_dev(idev);
2778 /* Step 2: clear hash table */
2779 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2780 struct hlist_head *h = &inet6_addr_lst[i];
2781 struct hlist_node *n;
2783 spin_lock_bh(&addrconf_hash_lock);
2784 restart:
2785 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2786 if (ifa->idev == idev) {
2787 hlist_del_init_rcu(&ifa->addr_lst);
2788 addrconf_del_timer(ifa);
2789 goto restart;
2792 spin_unlock_bh(&addrconf_hash_lock);
2795 write_lock_bh(&idev->lock);
2797 /* Step 2: clear flags for stateless addrconf */
2798 if (!how)
2799 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2801 #ifdef CONFIG_IPV6_PRIVACY
2802 if (how && del_timer(&idev->regen_timer))
2803 in6_dev_put(idev);
2805 /* Step 3: clear tempaddr list */
2806 while (!list_empty(&idev->tempaddr_list)) {
2807 ifa = list_first_entry(&idev->tempaddr_list,
2808 struct inet6_ifaddr, tmp_list);
2809 list_del(&ifa->tmp_list);
2810 write_unlock_bh(&idev->lock);
2811 spin_lock_bh(&ifa->lock);
2813 if (ifa->ifpub) {
2814 in6_ifa_put(ifa->ifpub);
2815 ifa->ifpub = NULL;
2817 spin_unlock_bh(&ifa->lock);
2818 in6_ifa_put(ifa);
2819 write_lock_bh(&idev->lock);
2821 #endif
2823 while (!list_empty(&idev->addr_list)) {
2824 ifa = list_first_entry(&idev->addr_list,
2825 struct inet6_ifaddr, if_list);
2826 addrconf_del_timer(ifa);
2828 list_del(&ifa->if_list);
2830 write_unlock_bh(&idev->lock);
2832 spin_lock_bh(&ifa->state_lock);
2833 state = ifa->state;
2834 ifa->state = INET6_IFADDR_STATE_DEAD;
2835 spin_unlock_bh(&ifa->state_lock);
2837 if (state != INET6_IFADDR_STATE_DEAD) {
2838 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2839 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2841 in6_ifa_put(ifa);
2843 write_lock_bh(&idev->lock);
2846 write_unlock_bh(&idev->lock);
2848 /* Step 5: Discard multicast list */
2849 if (how)
2850 ipv6_mc_destroy_dev(idev);
2851 else
2852 ipv6_mc_down(idev);
2854 idev->tstamp = jiffies;
2856 /* Last: Shot the device (if unregistered) */
2857 if (how) {
2858 addrconf_sysctl_unregister(idev);
2859 neigh_parms_release(&nd_tbl, idev->nd_parms);
2860 neigh_ifdown(&nd_tbl, dev);
2861 in6_dev_put(idev);
2863 return 0;
2866 static void addrconf_rs_timer(unsigned long data)
2868 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2869 struct inet6_dev *idev = ifp->idev;
2871 read_lock(&idev->lock);
2872 if (idev->dead || !(idev->if_flags & IF_READY))
2873 goto out;
2875 if (idev->cnf.forwarding)
2876 goto out;
2878 /* Announcement received after solicitation was sent */
2879 if (idev->if_flags & IF_RA_RCVD)
2880 goto out;
2882 spin_lock(&ifp->lock);
2883 if (ifp->probes++ < idev->cnf.rtr_solicits) {
2884 /* The wait after the last probe can be shorter */
2885 addrconf_mod_timer(ifp, AC_RS,
2886 (ifp->probes == idev->cnf.rtr_solicits) ?
2887 idev->cnf.rtr_solicit_delay :
2888 idev->cnf.rtr_solicit_interval);
2889 spin_unlock(&ifp->lock);
2891 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2892 } else {
2893 spin_unlock(&ifp->lock);
2895 * Note: we do not support deprecated "all on-link"
2896 * assumption any longer.
2898 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2899 idev->dev->name);
2902 out:
2903 read_unlock(&idev->lock);
2904 in6_ifa_put(ifp);
2908 * Duplicate Address Detection
2910 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2912 unsigned long rand_num;
2913 struct inet6_dev *idev = ifp->idev;
2915 if (ifp->flags & IFA_F_OPTIMISTIC)
2916 rand_num = 0;
2917 else
2918 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2920 ifp->probes = idev->cnf.dad_transmits;
2921 addrconf_mod_timer(ifp, AC_DAD, rand_num);
2924 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2926 struct inet6_dev *idev = ifp->idev;
2927 struct net_device *dev = idev->dev;
2929 addrconf_join_solict(dev, &ifp->addr);
2931 net_srandom(ifp->addr.s6_addr32[3]);
2933 read_lock_bh(&idev->lock);
2934 spin_lock(&ifp->lock);
2935 if (ifp->state == INET6_IFADDR_STATE_DEAD)
2936 goto out;
2938 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2939 idev->cnf.accept_dad < 1 ||
2940 !(ifp->flags&IFA_F_TENTATIVE) ||
2941 ifp->flags & IFA_F_NODAD) {
2942 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2943 spin_unlock(&ifp->lock);
2944 read_unlock_bh(&idev->lock);
2946 addrconf_dad_completed(ifp);
2947 return;
2950 if (!(idev->if_flags & IF_READY)) {
2951 spin_unlock(&ifp->lock);
2952 read_unlock_bh(&idev->lock);
2954 * If the device is not ready:
2955 * - keep it tentative if it is a permanent address.
2956 * - otherwise, kill it.
2958 in6_ifa_hold(ifp);
2959 addrconf_dad_stop(ifp, 0);
2960 return;
2964 * Optimistic nodes can start receiving
2965 * Frames right away
2967 if (ifp->flags & IFA_F_OPTIMISTIC)
2968 ip6_ins_rt(ifp->rt);
2970 addrconf_dad_kick(ifp);
2971 out:
2972 spin_unlock(&ifp->lock);
2973 read_unlock_bh(&idev->lock);
2976 static void addrconf_dad_timer(unsigned long data)
2978 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2979 struct inet6_dev *idev = ifp->idev;
2980 struct in6_addr mcaddr;
2982 if (!ifp->probes && addrconf_dad_end(ifp))
2983 goto out;
2985 read_lock(&idev->lock);
2986 if (idev->dead || !(idev->if_flags & IF_READY)) {
2987 read_unlock(&idev->lock);
2988 goto out;
2991 spin_lock(&ifp->lock);
2992 if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2993 spin_unlock(&ifp->lock);
2994 read_unlock(&idev->lock);
2995 goto out;
2998 if (ifp->probes == 0) {
3000 * DAD was successful
3003 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3004 spin_unlock(&ifp->lock);
3005 read_unlock(&idev->lock);
3007 addrconf_dad_completed(ifp);
3009 goto out;
3012 ifp->probes--;
3013 addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3014 spin_unlock(&ifp->lock);
3015 read_unlock(&idev->lock);
3017 /* send a neighbour solicitation for our addr */
3018 addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3019 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3020 out:
3021 in6_ifa_put(ifp);
3024 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3026 struct net_device *dev = ifp->idev->dev;
3029 * Configure the address for reception. Now it is valid.
3032 ipv6_ifa_notify(RTM_NEWADDR, ifp);
3034 /* If added prefix is link local and we are prepared to process
3035 router advertisements, start sending router solicitations.
3038 if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3039 ifp->idev->cnf.accept_ra == 2) &&
3040 ifp->idev->cnf.rtr_solicits > 0 &&
3041 (dev->flags&IFF_LOOPBACK) == 0 &&
3042 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3044 * If a host as already performed a random delay
3045 * [...] as part of DAD [...] there is no need
3046 * to delay again before sending the first RS
3048 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3050 spin_lock_bh(&ifp->lock);
3051 ifp->probes = 1;
3052 ifp->idev->if_flags |= IF_RS_SENT;
3053 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3054 spin_unlock_bh(&ifp->lock);
3058 static void addrconf_dad_run(struct inet6_dev *idev)
3060 struct inet6_ifaddr *ifp;
3062 read_lock_bh(&idev->lock);
3063 list_for_each_entry(ifp, &idev->addr_list, if_list) {
3064 spin_lock(&ifp->lock);
3065 if (ifp->flags & IFA_F_TENTATIVE &&
3066 ifp->state == INET6_IFADDR_STATE_DAD)
3067 addrconf_dad_kick(ifp);
3068 spin_unlock(&ifp->lock);
3070 read_unlock_bh(&idev->lock);
3073 #ifdef CONFIG_PROC_FS
3074 struct if6_iter_state {
3075 struct seq_net_private p;
3076 int bucket;
3077 int offset;
3080 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3082 struct inet6_ifaddr *ifa = NULL;
3083 struct if6_iter_state *state = seq->private;
3084 struct net *net = seq_file_net(seq);
3085 int p = 0;
3087 /* initial bucket if pos is 0 */
3088 if (pos == 0) {
3089 state->bucket = 0;
3090 state->offset = 0;
3093 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3094 struct hlist_node *n;
3095 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3096 addr_lst) {
3097 /* sync with offset */
3098 if (p < state->offset) {
3099 p++;
3100 continue;
3102 state->offset++;
3103 if (net_eq(dev_net(ifa->idev->dev), net))
3104 return ifa;
3107 /* prepare for next bucket */
3108 state->offset = 0;
3109 p = 0;
3111 return NULL;
3114 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3115 struct inet6_ifaddr *ifa)
3117 struct if6_iter_state *state = seq->private;
3118 struct net *net = seq_file_net(seq);
3119 struct hlist_node *n = &ifa->addr_lst;
3121 hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3122 state->offset++;
3123 if (net_eq(dev_net(ifa->idev->dev), net))
3124 return ifa;
3127 while (++state->bucket < IN6_ADDR_HSIZE) {
3128 state->offset = 0;
3129 hlist_for_each_entry_rcu_bh(ifa, n,
3130 &inet6_addr_lst[state->bucket], addr_lst) {
3131 state->offset++;
3132 if (net_eq(dev_net(ifa->idev->dev), net))
3133 return ifa;
3137 return NULL;
3140 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3141 __acquires(rcu_bh)
3143 rcu_read_lock_bh();
3144 return if6_get_first(seq, *pos);
3147 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3149 struct inet6_ifaddr *ifa;
3151 ifa = if6_get_next(seq, v);
3152 ++*pos;
3153 return ifa;
3156 static void if6_seq_stop(struct seq_file *seq, void *v)
3157 __releases(rcu_bh)
3159 rcu_read_unlock_bh();
3162 static int if6_seq_show(struct seq_file *seq, void *v)
3164 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3165 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3166 &ifp->addr,
3167 ifp->idev->dev->ifindex,
3168 ifp->prefix_len,
3169 ifp->scope,
3170 ifp->flags,
3171 ifp->idev->dev->name);
3172 return 0;
3175 static const struct seq_operations if6_seq_ops = {
3176 .start = if6_seq_start,
3177 .next = if6_seq_next,
3178 .show = if6_seq_show,
3179 .stop = if6_seq_stop,
3182 static int if6_seq_open(struct inode *inode, struct file *file)
3184 return seq_open_net(inode, file, &if6_seq_ops,
3185 sizeof(struct if6_iter_state));
3188 static const struct file_operations if6_fops = {
3189 .owner = THIS_MODULE,
3190 .open = if6_seq_open,
3191 .read = seq_read,
3192 .llseek = seq_lseek,
3193 .release = seq_release_net,
3196 static int __net_init if6_proc_net_init(struct net *net)
3198 if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3199 return -ENOMEM;
3200 return 0;
3203 static void __net_exit if6_proc_net_exit(struct net *net)
3205 proc_net_remove(net, "if_inet6");
3208 static struct pernet_operations if6_proc_net_ops = {
3209 .init = if6_proc_net_init,
3210 .exit = if6_proc_net_exit,
3213 int __init if6_proc_init(void)
3215 return register_pernet_subsys(&if6_proc_net_ops);
3218 void if6_proc_exit(void)
3220 unregister_pernet_subsys(&if6_proc_net_ops);
3222 #endif /* CONFIG_PROC_FS */
3224 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3225 /* Check if address is a home address configured on any interface. */
3226 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3228 int ret = 0;
3229 struct inet6_ifaddr *ifp = NULL;
3230 struct hlist_node *n;
3231 unsigned int hash = ipv6_addr_hash(addr);
3233 rcu_read_lock_bh();
3234 hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3235 if (!net_eq(dev_net(ifp->idev->dev), net))
3236 continue;
3237 if (ipv6_addr_equal(&ifp->addr, addr) &&
3238 (ifp->flags & IFA_F_HOMEADDRESS)) {
3239 ret = 1;
3240 break;
3243 rcu_read_unlock_bh();
3244 return ret;
3246 #endif
3249 * Periodic address status verification
3252 static void addrconf_verify(unsigned long foo)
3254 unsigned long now, next, next_sec, next_sched;
3255 struct inet6_ifaddr *ifp;
3256 struct hlist_node *node;
3257 int i;
3259 rcu_read_lock_bh();
3260 spin_lock(&addrconf_verify_lock);
3261 now = jiffies;
3262 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3264 del_timer(&addr_chk_timer);
3266 for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3267 restart:
3268 hlist_for_each_entry_rcu_bh(ifp, node,
3269 &inet6_addr_lst[i], addr_lst) {
3270 unsigned long age;
3272 if (ifp->flags & IFA_F_PERMANENT)
3273 continue;
3275 spin_lock(&ifp->lock);
3276 /* We try to batch several events at once. */
3277 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3279 if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3280 age >= ifp->valid_lft) {
3281 spin_unlock(&ifp->lock);
3282 in6_ifa_hold(ifp);
3283 ipv6_del_addr(ifp);
3284 goto restart;
3285 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3286 spin_unlock(&ifp->lock);
3287 continue;
3288 } else if (age >= ifp->prefered_lft) {
3289 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3290 int deprecate = 0;
3292 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3293 deprecate = 1;
3294 ifp->flags |= IFA_F_DEPRECATED;
3297 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3298 next = ifp->tstamp + ifp->valid_lft * HZ;
3300 spin_unlock(&ifp->lock);
3302 if (deprecate) {
3303 in6_ifa_hold(ifp);
3305 ipv6_ifa_notify(0, ifp);
3306 in6_ifa_put(ifp);
3307 goto restart;
3309 #ifdef CONFIG_IPV6_PRIVACY
3310 } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3311 !(ifp->flags&IFA_F_TENTATIVE)) {
3312 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3313 ifp->idev->cnf.dad_transmits *
3314 ifp->idev->nd_parms->retrans_time / HZ;
3316 if (age >= ifp->prefered_lft - regen_advance) {
3317 struct inet6_ifaddr *ifpub = ifp->ifpub;
3318 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3319 next = ifp->tstamp + ifp->prefered_lft * HZ;
3320 if (!ifp->regen_count && ifpub) {
3321 ifp->regen_count++;
3322 in6_ifa_hold(ifp);
3323 in6_ifa_hold(ifpub);
3324 spin_unlock(&ifp->lock);
3326 spin_lock(&ifpub->lock);
3327 ifpub->regen_count = 0;
3328 spin_unlock(&ifpub->lock);
3329 ipv6_create_tempaddr(ifpub, ifp);
3330 in6_ifa_put(ifpub);
3331 in6_ifa_put(ifp);
3332 goto restart;
3334 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3335 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3336 spin_unlock(&ifp->lock);
3337 #endif
3338 } else {
3339 /* ifp->prefered_lft <= ifp->valid_lft */
3340 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3341 next = ifp->tstamp + ifp->prefered_lft * HZ;
3342 spin_unlock(&ifp->lock);
3347 next_sec = round_jiffies_up(next);
3348 next_sched = next;
3350 /* If rounded timeout is accurate enough, accept it. */
3351 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3352 next_sched = next_sec;
3354 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3355 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3356 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3358 ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3359 now, next, next_sec, next_sched));
3361 addr_chk_timer.expires = next_sched;
3362 add_timer(&addr_chk_timer);
3363 spin_unlock(&addrconf_verify_lock);
3364 rcu_read_unlock_bh();
3367 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3369 struct in6_addr *pfx = NULL;
3371 if (addr)
3372 pfx = nla_data(addr);
3374 if (local) {
3375 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3376 pfx = NULL;
3377 else
3378 pfx = nla_data(local);
3381 return pfx;
3384 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3385 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) },
3386 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) },
3387 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) },
3390 static int
3391 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3393 struct net *net = sock_net(skb->sk);
3394 struct ifaddrmsg *ifm;
3395 struct nlattr *tb[IFA_MAX+1];
3396 struct in6_addr *pfx;
3397 int err;
3399 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3400 if (err < 0)
3401 return err;
3403 ifm = nlmsg_data(nlh);
3404 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3405 if (pfx == NULL)
3406 return -EINVAL;
3408 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3411 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3412 u32 prefered_lft, u32 valid_lft)
3414 u32 flags;
3415 clock_t expires;
3416 unsigned long timeout;
3418 if (!valid_lft || (prefered_lft > valid_lft))
3419 return -EINVAL;
3421 timeout = addrconf_timeout_fixup(valid_lft, HZ);
3422 if (addrconf_finite_timeout(timeout)) {
3423 expires = jiffies_to_clock_t(timeout * HZ);
3424 valid_lft = timeout;
3425 flags = RTF_EXPIRES;
3426 } else {
3427 expires = 0;
3428 flags = 0;
3429 ifa_flags |= IFA_F_PERMANENT;
3432 timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3433 if (addrconf_finite_timeout(timeout)) {
3434 if (timeout == 0)
3435 ifa_flags |= IFA_F_DEPRECATED;
3436 prefered_lft = timeout;
3439 spin_lock_bh(&ifp->lock);
3440 ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3441 ifp->tstamp = jiffies;
3442 ifp->valid_lft = valid_lft;
3443 ifp->prefered_lft = prefered_lft;
3445 spin_unlock_bh(&ifp->lock);
3446 if (!(ifp->flags&IFA_F_TENTATIVE))
3447 ipv6_ifa_notify(0, ifp);
3449 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3450 expires, flags);
3451 addrconf_verify(0);
3453 return 0;
3456 static int
3457 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3459 struct net *net = sock_net(skb->sk);
3460 struct ifaddrmsg *ifm;
3461 struct nlattr *tb[IFA_MAX+1];
3462 struct in6_addr *pfx;
3463 struct inet6_ifaddr *ifa;
3464 struct net_device *dev;
3465 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3466 u8 ifa_flags;
3467 int err;
3469 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3470 if (err < 0)
3471 return err;
3473 ifm = nlmsg_data(nlh);
3474 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3475 if (pfx == NULL)
3476 return -EINVAL;
3478 if (tb[IFA_CACHEINFO]) {
3479 struct ifa_cacheinfo *ci;
3481 ci = nla_data(tb[IFA_CACHEINFO]);
3482 valid_lft = ci->ifa_valid;
3483 preferred_lft = ci->ifa_prefered;
3484 } else {
3485 preferred_lft = INFINITY_LIFE_TIME;
3486 valid_lft = INFINITY_LIFE_TIME;
3489 dev = __dev_get_by_index(net, ifm->ifa_index);
3490 if (dev == NULL)
3491 return -ENODEV;
3493 /* We ignore other flags so far. */
3494 ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3496 ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3497 if (ifa == NULL) {
3499 * It would be best to check for !NLM_F_CREATE here but
3500 * userspace alreay relies on not having to provide this.
3502 return inet6_addr_add(net, ifm->ifa_index, pfx,
3503 ifm->ifa_prefixlen, ifa_flags,
3504 preferred_lft, valid_lft);
3507 if (nlh->nlmsg_flags & NLM_F_EXCL ||
3508 !(nlh->nlmsg_flags & NLM_F_REPLACE))
3509 err = -EEXIST;
3510 else
3511 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3513 in6_ifa_put(ifa);
3515 return err;
3518 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3519 u8 scope, int ifindex)
3521 struct ifaddrmsg *ifm;
3523 ifm = nlmsg_data(nlh);
3524 ifm->ifa_family = AF_INET6;
3525 ifm->ifa_prefixlen = prefixlen;
3526 ifm->ifa_flags = flags;
3527 ifm->ifa_scope = scope;
3528 ifm->ifa_index = ifindex;
3531 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3532 unsigned long tstamp, u32 preferred, u32 valid)
3534 struct ifa_cacheinfo ci;
3536 ci.cstamp = cstamp_delta(cstamp);
3537 ci.tstamp = cstamp_delta(tstamp);
3538 ci.ifa_prefered = preferred;
3539 ci.ifa_valid = valid;
3541 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3544 static inline int rt_scope(int ifa_scope)
3546 if (ifa_scope & IFA_HOST)
3547 return RT_SCOPE_HOST;
3548 else if (ifa_scope & IFA_LINK)
3549 return RT_SCOPE_LINK;
3550 else if (ifa_scope & IFA_SITE)
3551 return RT_SCOPE_SITE;
3552 else
3553 return RT_SCOPE_UNIVERSE;
3556 static inline int inet6_ifaddr_msgsize(void)
3558 return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3559 + nla_total_size(16) /* IFA_ADDRESS */
3560 + nla_total_size(sizeof(struct ifa_cacheinfo));
3563 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3564 u32 pid, u32 seq, int event, unsigned int flags)
3566 struct nlmsghdr *nlh;
3567 u32 preferred, valid;
3569 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3570 if (nlh == NULL)
3571 return -EMSGSIZE;
3573 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3574 ifa->idev->dev->ifindex);
3576 if (!(ifa->flags&IFA_F_PERMANENT)) {
3577 preferred = ifa->prefered_lft;
3578 valid = ifa->valid_lft;
3579 if (preferred != INFINITY_LIFE_TIME) {
3580 long tval = (jiffies - ifa->tstamp)/HZ;
3581 if (preferred > tval)
3582 preferred -= tval;
3583 else
3584 preferred = 0;
3585 if (valid != INFINITY_LIFE_TIME) {
3586 if (valid > tval)
3587 valid -= tval;
3588 else
3589 valid = 0;
3592 } else {
3593 preferred = INFINITY_LIFE_TIME;
3594 valid = INFINITY_LIFE_TIME;
3597 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3598 put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3599 nlmsg_cancel(skb, nlh);
3600 return -EMSGSIZE;
3603 return nlmsg_end(skb, nlh);
3606 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3607 u32 pid, u32 seq, int event, u16 flags)
3609 struct nlmsghdr *nlh;
3610 u8 scope = RT_SCOPE_UNIVERSE;
3611 int ifindex = ifmca->idev->dev->ifindex;
3613 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3614 scope = RT_SCOPE_SITE;
3616 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3617 if (nlh == NULL)
3618 return -EMSGSIZE;
3620 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3621 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3622 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3623 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3624 nlmsg_cancel(skb, nlh);
3625 return -EMSGSIZE;
3628 return nlmsg_end(skb, nlh);
3631 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3632 u32 pid, u32 seq, int event, unsigned int flags)
3634 struct nlmsghdr *nlh;
3635 u8 scope = RT_SCOPE_UNIVERSE;
3636 int ifindex = ifaca->aca_idev->dev->ifindex;
3638 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3639 scope = RT_SCOPE_SITE;
3641 nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3642 if (nlh == NULL)
3643 return -EMSGSIZE;
3645 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3646 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3647 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3648 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3649 nlmsg_cancel(skb, nlh);
3650 return -EMSGSIZE;
3653 return nlmsg_end(skb, nlh);
3656 enum addr_type_t {
3657 UNICAST_ADDR,
3658 MULTICAST_ADDR,
3659 ANYCAST_ADDR,
3662 /* called with rcu_read_lock() */
3663 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3664 struct netlink_callback *cb, enum addr_type_t type,
3665 int s_ip_idx, int *p_ip_idx)
3667 struct ifmcaddr6 *ifmca;
3668 struct ifacaddr6 *ifaca;
3669 int err = 1;
3670 int ip_idx = *p_ip_idx;
3672 read_lock_bh(&idev->lock);
3673 switch (type) {
3674 case UNICAST_ADDR: {
3675 struct inet6_ifaddr *ifa;
3677 /* unicast address incl. temp addr */
3678 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3679 if (++ip_idx < s_ip_idx)
3680 continue;
3681 err = inet6_fill_ifaddr(skb, ifa,
3682 NETLINK_CB(cb->skb).pid,
3683 cb->nlh->nlmsg_seq,
3684 RTM_NEWADDR,
3685 NLM_F_MULTI);
3686 if (err <= 0)
3687 break;
3689 break;
3691 case MULTICAST_ADDR:
3692 /* multicast address */
3693 for (ifmca = idev->mc_list; ifmca;
3694 ifmca = ifmca->next, ip_idx++) {
3695 if (ip_idx < s_ip_idx)
3696 continue;
3697 err = inet6_fill_ifmcaddr(skb, ifmca,
3698 NETLINK_CB(cb->skb).pid,
3699 cb->nlh->nlmsg_seq,
3700 RTM_GETMULTICAST,
3701 NLM_F_MULTI);
3702 if (err <= 0)
3703 break;
3705 break;
3706 case ANYCAST_ADDR:
3707 /* anycast address */
3708 for (ifaca = idev->ac_list; ifaca;
3709 ifaca = ifaca->aca_next, ip_idx++) {
3710 if (ip_idx < s_ip_idx)
3711 continue;
3712 err = inet6_fill_ifacaddr(skb, ifaca,
3713 NETLINK_CB(cb->skb).pid,
3714 cb->nlh->nlmsg_seq,
3715 RTM_GETANYCAST,
3716 NLM_F_MULTI);
3717 if (err <= 0)
3718 break;
3720 break;
3721 default:
3722 break;
3724 read_unlock_bh(&idev->lock);
3725 *p_ip_idx = ip_idx;
3726 return err;
3729 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3730 enum addr_type_t type)
3732 struct net *net = sock_net(skb->sk);
3733 int h, s_h;
3734 int idx, ip_idx;
3735 int s_idx, s_ip_idx;
3736 struct net_device *dev;
3737 struct inet6_dev *idev;
3738 struct hlist_head *head;
3739 struct hlist_node *node;
3741 s_h = cb->args[0];
3742 s_idx = idx = cb->args[1];
3743 s_ip_idx = ip_idx = cb->args[2];
3745 rcu_read_lock();
3746 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3747 idx = 0;
3748 head = &net->dev_index_head[h];
3749 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3750 if (idx < s_idx)
3751 goto cont;
3752 if (h > s_h || idx > s_idx)
3753 s_ip_idx = 0;
3754 ip_idx = 0;
3755 idev = __in6_dev_get(dev);
3756 if (!idev)
3757 goto cont;
3759 if (in6_dump_addrs(idev, skb, cb, type,
3760 s_ip_idx, &ip_idx) <= 0)
3761 goto done;
3762 cont:
3763 idx++;
3766 done:
3767 rcu_read_unlock();
3768 cb->args[0] = h;
3769 cb->args[1] = idx;
3770 cb->args[2] = ip_idx;
3772 return skb->len;
3775 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3777 enum addr_type_t type = UNICAST_ADDR;
3779 return inet6_dump_addr(skb, cb, type);
3782 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3784 enum addr_type_t type = MULTICAST_ADDR;
3786 return inet6_dump_addr(skb, cb, type);
3790 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3792 enum addr_type_t type = ANYCAST_ADDR;
3794 return inet6_dump_addr(skb, cb, type);
3797 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3798 void *arg)
3800 struct net *net = sock_net(in_skb->sk);
3801 struct ifaddrmsg *ifm;
3802 struct nlattr *tb[IFA_MAX+1];
3803 struct in6_addr *addr = NULL;
3804 struct net_device *dev = NULL;
3805 struct inet6_ifaddr *ifa;
3806 struct sk_buff *skb;
3807 int err;
3809 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3810 if (err < 0)
3811 goto errout;
3813 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3814 if (addr == NULL) {
3815 err = -EINVAL;
3816 goto errout;
3819 ifm = nlmsg_data(nlh);
3820 if (ifm->ifa_index)
3821 dev = __dev_get_by_index(net, ifm->ifa_index);
3823 ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3824 if (!ifa) {
3825 err = -EADDRNOTAVAIL;
3826 goto errout;
3829 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3830 if (!skb) {
3831 err = -ENOBUFS;
3832 goto errout_ifa;
3835 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3836 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3837 if (err < 0) {
3838 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3839 WARN_ON(err == -EMSGSIZE);
3840 kfree_skb(skb);
3841 goto errout_ifa;
3843 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3844 errout_ifa:
3845 in6_ifa_put(ifa);
3846 errout:
3847 return err;
3850 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3852 struct sk_buff *skb;
3853 struct net *net = dev_net(ifa->idev->dev);
3854 int err = -ENOBUFS;
3856 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3857 if (skb == NULL)
3858 goto errout;
3860 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3861 if (err < 0) {
3862 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3863 WARN_ON(err == -EMSGSIZE);
3864 kfree_skb(skb);
3865 goto errout;
3867 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3868 return;
3869 errout:
3870 if (err < 0)
3871 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3874 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3875 __s32 *array, int bytes)
3877 BUG_ON(bytes < (DEVCONF_MAX * 4));
3879 memset(array, 0, bytes);
3880 array[DEVCONF_FORWARDING] = cnf->forwarding;
3881 array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3882 array[DEVCONF_MTU6] = cnf->mtu6;
3883 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3884 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3885 array[DEVCONF_AUTOCONF] = cnf->autoconf;
3886 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3887 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3888 array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3889 jiffies_to_msecs(cnf->rtr_solicit_interval);
3890 array[DEVCONF_RTR_SOLICIT_DELAY] =
3891 jiffies_to_msecs(cnf->rtr_solicit_delay);
3892 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3893 #ifdef CONFIG_IPV6_PRIVACY
3894 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3895 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3896 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3897 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3898 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3899 #endif
3900 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3901 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3902 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3903 #ifdef CONFIG_IPV6_ROUTER_PREF
3904 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3905 array[DEVCONF_RTR_PROBE_INTERVAL] =
3906 jiffies_to_msecs(cnf->rtr_probe_interval);
3907 #ifdef CONFIG_IPV6_ROUTE_INFO
3908 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3909 #endif
3910 #endif
3911 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3912 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3913 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3914 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3915 #endif
3916 #ifdef CONFIG_IPV6_MROUTE
3917 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3918 #endif
3919 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3920 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3921 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3924 static inline size_t inet6_ifla6_size(void)
3926 return nla_total_size(4) /* IFLA_INET6_FLAGS */
3927 + nla_total_size(sizeof(struct ifla_cacheinfo))
3928 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3929 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3930 + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3933 static inline size_t inet6_if_nlmsg_size(void)
3935 return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3936 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3937 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3938 + nla_total_size(4) /* IFLA_MTU */
3939 + nla_total_size(4) /* IFLA_LINK */
3940 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3943 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3944 int items, int bytes)
3946 int i;
3947 int pad = bytes - sizeof(u64) * items;
3948 BUG_ON(pad < 0);
3950 /* Use put_unaligned() because stats may not be aligned for u64. */
3951 put_unaligned(items, &stats[0]);
3952 for (i = 1; i < items; i++)
3953 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3955 memset(&stats[items], 0, pad);
3958 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3959 int items, int bytes, size_t syncpoff)
3961 int i;
3962 int pad = bytes - sizeof(u64) * items;
3963 BUG_ON(pad < 0);
3965 /* Use put_unaligned() because stats may not be aligned for u64. */
3966 put_unaligned(items, &stats[0]);
3967 for (i = 1; i < items; i++)
3968 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3970 memset(&stats[items], 0, pad);
3973 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3974 int bytes)
3976 switch (attrtype) {
3977 case IFLA_INET6_STATS:
3978 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3979 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3980 break;
3981 case IFLA_INET6_ICMP6STATS:
3982 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3983 break;
3987 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3989 struct nlattr *nla;
3990 struct ifla_cacheinfo ci;
3992 NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3994 ci.max_reasm_len = IPV6_MAXPLEN;
3995 ci.tstamp = cstamp_delta(idev->tstamp);
3996 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3997 ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3998 NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
4000 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4001 if (nla == NULL)
4002 goto nla_put_failure;
4003 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4005 /* XXX - MC not implemented */
4007 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4008 if (nla == NULL)
4009 goto nla_put_failure;
4010 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4012 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4013 if (nla == NULL)
4014 goto nla_put_failure;
4015 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4017 return 0;
4019 nla_put_failure:
4020 return -EMSGSIZE;
4023 static size_t inet6_get_link_af_size(const struct net_device *dev)
4025 if (!__in6_dev_get(dev))
4026 return 0;
4028 return inet6_ifla6_size();
4031 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4033 struct inet6_dev *idev = __in6_dev_get(dev);
4035 if (!idev)
4036 return -ENODATA;
4038 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4039 return -EMSGSIZE;
4041 return 0;
4044 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4045 u32 pid, u32 seq, int event, unsigned int flags)
4047 struct net_device *dev = idev->dev;
4048 struct ifinfomsg *hdr;
4049 struct nlmsghdr *nlh;
4050 void *protoinfo;
4052 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4053 if (nlh == NULL)
4054 return -EMSGSIZE;
4056 hdr = nlmsg_data(nlh);
4057 hdr->ifi_family = AF_INET6;
4058 hdr->__ifi_pad = 0;
4059 hdr->ifi_type = dev->type;
4060 hdr->ifi_index = dev->ifindex;
4061 hdr->ifi_flags = dev_get_flags(dev);
4062 hdr->ifi_change = 0;
4064 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4066 if (dev->addr_len)
4067 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4069 NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4070 if (dev->ifindex != dev->iflink)
4071 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4073 protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4074 if (protoinfo == NULL)
4075 goto nla_put_failure;
4077 if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4078 goto nla_put_failure;
4080 nla_nest_end(skb, protoinfo);
4081 return nlmsg_end(skb, nlh);
4083 nla_put_failure:
4084 nlmsg_cancel(skb, nlh);
4085 return -EMSGSIZE;
4088 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4090 struct net *net = sock_net(skb->sk);
4091 int h, s_h;
4092 int idx = 0, s_idx;
4093 struct net_device *dev;
4094 struct inet6_dev *idev;
4095 struct hlist_head *head;
4096 struct hlist_node *node;
4098 s_h = cb->args[0];
4099 s_idx = cb->args[1];
4101 rcu_read_lock();
4102 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4103 idx = 0;
4104 head = &net->dev_index_head[h];
4105 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4106 if (idx < s_idx)
4107 goto cont;
4108 idev = __in6_dev_get(dev);
4109 if (!idev)
4110 goto cont;
4111 if (inet6_fill_ifinfo(skb, idev,
4112 NETLINK_CB(cb->skb).pid,
4113 cb->nlh->nlmsg_seq,
4114 RTM_NEWLINK, NLM_F_MULTI) <= 0)
4115 goto out;
4116 cont:
4117 idx++;
4120 out:
4121 rcu_read_unlock();
4122 cb->args[1] = idx;
4123 cb->args[0] = h;
4125 return skb->len;
4128 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4130 struct sk_buff *skb;
4131 struct net *net = dev_net(idev->dev);
4132 int err = -ENOBUFS;
4134 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4135 if (skb == NULL)
4136 goto errout;
4138 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4139 if (err < 0) {
4140 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4141 WARN_ON(err == -EMSGSIZE);
4142 kfree_skb(skb);
4143 goto errout;
4145 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4146 return;
4147 errout:
4148 if (err < 0)
4149 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4152 static inline size_t inet6_prefix_nlmsg_size(void)
4154 return NLMSG_ALIGN(sizeof(struct prefixmsg))
4155 + nla_total_size(sizeof(struct in6_addr))
4156 + nla_total_size(sizeof(struct prefix_cacheinfo));
4159 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4160 struct prefix_info *pinfo, u32 pid, u32 seq,
4161 int event, unsigned int flags)
4163 struct prefixmsg *pmsg;
4164 struct nlmsghdr *nlh;
4165 struct prefix_cacheinfo ci;
4167 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4168 if (nlh == NULL)
4169 return -EMSGSIZE;
4171 pmsg = nlmsg_data(nlh);
4172 pmsg->prefix_family = AF_INET6;
4173 pmsg->prefix_pad1 = 0;
4174 pmsg->prefix_pad2 = 0;
4175 pmsg->prefix_ifindex = idev->dev->ifindex;
4176 pmsg->prefix_len = pinfo->prefix_len;
4177 pmsg->prefix_type = pinfo->type;
4178 pmsg->prefix_pad3 = 0;
4179 pmsg->prefix_flags = 0;
4180 if (pinfo->onlink)
4181 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4182 if (pinfo->autoconf)
4183 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4185 NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4187 ci.preferred_time = ntohl(pinfo->prefered);
4188 ci.valid_time = ntohl(pinfo->valid);
4189 NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4191 return nlmsg_end(skb, nlh);
4193 nla_put_failure:
4194 nlmsg_cancel(skb, nlh);
4195 return -EMSGSIZE;
4198 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4199 struct prefix_info *pinfo)
4201 struct sk_buff *skb;
4202 struct net *net = dev_net(idev->dev);
4203 int err = -ENOBUFS;
4205 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4206 if (skb == NULL)
4207 goto errout;
4209 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4210 if (err < 0) {
4211 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4212 WARN_ON(err == -EMSGSIZE);
4213 kfree_skb(skb);
4214 goto errout;
4216 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4217 return;
4218 errout:
4219 if (err < 0)
4220 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4223 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4225 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4227 switch (event) {
4228 case RTM_NEWADDR:
4230 * If the address was optimistic
4231 * we inserted the route at the start of
4232 * our DAD process, so we don't need
4233 * to do it again
4235 if (!(ifp->rt->rt6i_node))
4236 ip6_ins_rt(ifp->rt);
4237 if (ifp->idev->cnf.forwarding)
4238 addrconf_join_anycast(ifp);
4239 break;
4240 case RTM_DELADDR:
4241 if (ifp->idev->cnf.forwarding)
4242 addrconf_leave_anycast(ifp);
4243 addrconf_leave_solict(ifp->idev, &ifp->addr);
4244 dst_hold(&ifp->rt->dst);
4246 if (ip6_del_rt(ifp->rt))
4247 dst_free(&ifp->rt->dst);
4248 break;
4252 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4254 rcu_read_lock_bh();
4255 if (likely(ifp->idev->dead == 0))
4256 __ipv6_ifa_notify(event, ifp);
4257 rcu_read_unlock_bh();
4260 #ifdef CONFIG_SYSCTL
4262 static
4263 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4264 void __user *buffer, size_t *lenp, loff_t *ppos)
4266 int *valp = ctl->data;
4267 int val = *valp;
4268 loff_t pos = *ppos;
4269 ctl_table lctl;
4270 int ret;
4273 * ctl->data points to idev->cnf.forwarding, we should
4274 * not modify it until we get the rtnl lock.
4276 lctl = *ctl;
4277 lctl.data = &val;
4279 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4281 if (write)
4282 ret = addrconf_fixup_forwarding(ctl, valp, val);
4283 if (ret)
4284 *ppos = pos;
4285 return ret;
4288 static void dev_disable_change(struct inet6_dev *idev)
4290 if (!idev || !idev->dev)
4291 return;
4293 if (idev->cnf.disable_ipv6)
4294 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4295 else
4296 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4299 static void addrconf_disable_change(struct net *net, __s32 newf)
4301 struct net_device *dev;
4302 struct inet6_dev *idev;
4304 rcu_read_lock();
4305 for_each_netdev_rcu(net, dev) {
4306 idev = __in6_dev_get(dev);
4307 if (idev) {
4308 int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4309 idev->cnf.disable_ipv6 = newf;
4310 if (changed)
4311 dev_disable_change(idev);
4314 rcu_read_unlock();
4317 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4319 struct net *net;
4320 int old;
4322 if (!rtnl_trylock())
4323 return restart_syscall();
4325 net = (struct net *)table->extra2;
4326 old = *p;
4327 *p = newf;
4329 if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4330 rtnl_unlock();
4331 return 0;
4334 if (p == &net->ipv6.devconf_all->disable_ipv6) {
4335 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4336 addrconf_disable_change(net, newf);
4337 } else if ((!newf) ^ (!old))
4338 dev_disable_change((struct inet6_dev *)table->extra1);
4340 rtnl_unlock();
4341 return 0;
4344 static
4345 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4346 void __user *buffer, size_t *lenp, loff_t *ppos)
4348 int *valp = ctl->data;
4349 int val = *valp;
4350 loff_t pos = *ppos;
4351 ctl_table lctl;
4352 int ret;
4355 * ctl->data points to idev->cnf.disable_ipv6, we should
4356 * not modify it until we get the rtnl lock.
4358 lctl = *ctl;
4359 lctl.data = &val;
4361 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4363 if (write)
4364 ret = addrconf_disable_ipv6(ctl, valp, val);
4365 if (ret)
4366 *ppos = pos;
4367 return ret;
4370 static struct addrconf_sysctl_table
4372 struct ctl_table_header *sysctl_header;
4373 ctl_table addrconf_vars[DEVCONF_MAX+1];
4374 char *dev_name;
4375 } addrconf_sysctl __read_mostly = {
4376 .sysctl_header = NULL,
4377 .addrconf_vars = {
4379 .procname = "forwarding",
4380 .data = &ipv6_devconf.forwarding,
4381 .maxlen = sizeof(int),
4382 .mode = 0644,
4383 .proc_handler = addrconf_sysctl_forward,
4386 .procname = "hop_limit",
4387 .data = &ipv6_devconf.hop_limit,
4388 .maxlen = sizeof(int),
4389 .mode = 0644,
4390 .proc_handler = proc_dointvec,
4393 .procname = "mtu",
4394 .data = &ipv6_devconf.mtu6,
4395 .maxlen = sizeof(int),
4396 .mode = 0644,
4397 .proc_handler = proc_dointvec,
4400 .procname = "accept_ra",
4401 .data = &ipv6_devconf.accept_ra,
4402 .maxlen = sizeof(int),
4403 .mode = 0644,
4404 .proc_handler = proc_dointvec,
4407 .procname = "accept_redirects",
4408 .data = &ipv6_devconf.accept_redirects,
4409 .maxlen = sizeof(int),
4410 .mode = 0644,
4411 .proc_handler = proc_dointvec,
4414 .procname = "autoconf",
4415 .data = &ipv6_devconf.autoconf,
4416 .maxlen = sizeof(int),
4417 .mode = 0644,
4418 .proc_handler = proc_dointvec,
4421 .procname = "dad_transmits",
4422 .data = &ipv6_devconf.dad_transmits,
4423 .maxlen = sizeof(int),
4424 .mode = 0644,
4425 .proc_handler = proc_dointvec,
4428 .procname = "router_solicitations",
4429 .data = &ipv6_devconf.rtr_solicits,
4430 .maxlen = sizeof(int),
4431 .mode = 0644,
4432 .proc_handler = proc_dointvec,
4435 .procname = "router_solicitation_interval",
4436 .data = &ipv6_devconf.rtr_solicit_interval,
4437 .maxlen = sizeof(int),
4438 .mode = 0644,
4439 .proc_handler = proc_dointvec_jiffies,
4442 .procname = "router_solicitation_delay",
4443 .data = &ipv6_devconf.rtr_solicit_delay,
4444 .maxlen = sizeof(int),
4445 .mode = 0644,
4446 .proc_handler = proc_dointvec_jiffies,
4449 .procname = "force_mld_version",
4450 .data = &ipv6_devconf.force_mld_version,
4451 .maxlen = sizeof(int),
4452 .mode = 0644,
4453 .proc_handler = proc_dointvec,
4455 #ifdef CONFIG_IPV6_PRIVACY
4457 .procname = "use_tempaddr",
4458 .data = &ipv6_devconf.use_tempaddr,
4459 .maxlen = sizeof(int),
4460 .mode = 0644,
4461 .proc_handler = proc_dointvec,
4464 .procname = "temp_valid_lft",
4465 .data = &ipv6_devconf.temp_valid_lft,
4466 .maxlen = sizeof(int),
4467 .mode = 0644,
4468 .proc_handler = proc_dointvec,
4471 .procname = "temp_prefered_lft",
4472 .data = &ipv6_devconf.temp_prefered_lft,
4473 .maxlen = sizeof(int),
4474 .mode = 0644,
4475 .proc_handler = proc_dointvec,
4478 .procname = "regen_max_retry",
4479 .data = &ipv6_devconf.regen_max_retry,
4480 .maxlen = sizeof(int),
4481 .mode = 0644,
4482 .proc_handler = proc_dointvec,
4485 .procname = "max_desync_factor",
4486 .data = &ipv6_devconf.max_desync_factor,
4487 .maxlen = sizeof(int),
4488 .mode = 0644,
4489 .proc_handler = proc_dointvec,
4491 #endif
4493 .procname = "max_addresses",
4494 .data = &ipv6_devconf.max_addresses,
4495 .maxlen = sizeof(int),
4496 .mode = 0644,
4497 .proc_handler = proc_dointvec,
4500 .procname = "accept_ra_defrtr",
4501 .data = &ipv6_devconf.accept_ra_defrtr,
4502 .maxlen = sizeof(int),
4503 .mode = 0644,
4504 .proc_handler = proc_dointvec,
4507 .procname = "accept_ra_pinfo",
4508 .data = &ipv6_devconf.accept_ra_pinfo,
4509 .maxlen = sizeof(int),
4510 .mode = 0644,
4511 .proc_handler = proc_dointvec,
4513 #ifdef CONFIG_IPV6_ROUTER_PREF
4515 .procname = "accept_ra_rtr_pref",
4516 .data = &ipv6_devconf.accept_ra_rtr_pref,
4517 .maxlen = sizeof(int),
4518 .mode = 0644,
4519 .proc_handler = proc_dointvec,
4522 .procname = "router_probe_interval",
4523 .data = &ipv6_devconf.rtr_probe_interval,
4524 .maxlen = sizeof(int),
4525 .mode = 0644,
4526 .proc_handler = proc_dointvec_jiffies,
4528 #ifdef CONFIG_IPV6_ROUTE_INFO
4530 .procname = "accept_ra_rt_info_max_plen",
4531 .data = &ipv6_devconf.accept_ra_rt_info_max_plen,
4532 .maxlen = sizeof(int),
4533 .mode = 0644,
4534 .proc_handler = proc_dointvec,
4536 #endif
4537 #endif
4539 .procname = "proxy_ndp",
4540 .data = &ipv6_devconf.proxy_ndp,
4541 .maxlen = sizeof(int),
4542 .mode = 0644,
4543 .proc_handler = proc_dointvec,
4546 .procname = "accept_source_route",
4547 .data = &ipv6_devconf.accept_source_route,
4548 .maxlen = sizeof(int),
4549 .mode = 0644,
4550 .proc_handler = proc_dointvec,
4552 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4554 .procname = "optimistic_dad",
4555 .data = &ipv6_devconf.optimistic_dad,
4556 .maxlen = sizeof(int),
4557 .mode = 0644,
4558 .proc_handler = proc_dointvec,
4561 #endif
4562 #ifdef CONFIG_IPV6_MROUTE
4564 .procname = "mc_forwarding",
4565 .data = &ipv6_devconf.mc_forwarding,
4566 .maxlen = sizeof(int),
4567 .mode = 0444,
4568 .proc_handler = proc_dointvec,
4570 #endif
4572 .procname = "disable_ipv6",
4573 .data = &ipv6_devconf.disable_ipv6,
4574 .maxlen = sizeof(int),
4575 .mode = 0644,
4576 .proc_handler = addrconf_sysctl_disable,
4579 .procname = "accept_dad",
4580 .data = &ipv6_devconf.accept_dad,
4581 .maxlen = sizeof(int),
4582 .mode = 0644,
4583 .proc_handler = proc_dointvec,
4586 .procname = "force_tllao",
4587 .data = &ipv6_devconf.force_tllao,
4588 .maxlen = sizeof(int),
4589 .mode = 0644,
4590 .proc_handler = proc_dointvec
4593 /* sentinel */
4598 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4599 struct inet6_dev *idev, struct ipv6_devconf *p)
4601 int i;
4602 struct addrconf_sysctl_table *t;
4604 #define ADDRCONF_CTL_PATH_DEV 3
4606 struct ctl_path addrconf_ctl_path[] = {
4607 { .procname = "net", },
4608 { .procname = "ipv6", },
4609 { .procname = "conf", },
4610 { /* to be set */ },
4611 { },
4615 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4616 if (t == NULL)
4617 goto out;
4619 for (i = 0; t->addrconf_vars[i].data; i++) {
4620 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4621 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4622 t->addrconf_vars[i].extra2 = net;
4626 * Make a copy of dev_name, because '.procname' is regarded as const
4627 * by sysctl and we wouldn't want anyone to change it under our feet
4628 * (see SIOCSIFNAME).
4630 t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4631 if (!t->dev_name)
4632 goto free;
4634 addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4636 t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4637 t->addrconf_vars);
4638 if (t->sysctl_header == NULL)
4639 goto free_procname;
4641 p->sysctl = t;
4642 return 0;
4644 free_procname:
4645 kfree(t->dev_name);
4646 free:
4647 kfree(t);
4648 out:
4649 return -ENOBUFS;
4652 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4654 struct addrconf_sysctl_table *t;
4656 if (p->sysctl == NULL)
4657 return;
4659 t = p->sysctl;
4660 p->sysctl = NULL;
4661 unregister_net_sysctl_table(t->sysctl_header);
4662 kfree(t->dev_name);
4663 kfree(t);
4666 static void addrconf_sysctl_register(struct inet6_dev *idev)
4668 neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4669 &ndisc_ifinfo_sysctl_change);
4670 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4671 idev, &idev->cnf);
4674 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4676 __addrconf_sysctl_unregister(&idev->cnf);
4677 neigh_sysctl_unregister(idev->nd_parms);
4681 #endif
4683 static int __net_init addrconf_init_net(struct net *net)
4685 int err;
4686 struct ipv6_devconf *all, *dflt;
4688 err = -ENOMEM;
4689 all = &ipv6_devconf;
4690 dflt = &ipv6_devconf_dflt;
4692 if (!net_eq(net, &init_net)) {
4693 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4694 if (all == NULL)
4695 goto err_alloc_all;
4697 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4698 if (dflt == NULL)
4699 goto err_alloc_dflt;
4700 } else {
4701 /* these will be inherited by all namespaces */
4702 dflt->autoconf = ipv6_defaults.autoconf;
4703 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4706 net->ipv6.devconf_all = all;
4707 net->ipv6.devconf_dflt = dflt;
4709 #ifdef CONFIG_SYSCTL
4710 err = __addrconf_sysctl_register(net, "all", NULL, all);
4711 if (err < 0)
4712 goto err_reg_all;
4714 err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4715 if (err < 0)
4716 goto err_reg_dflt;
4717 #endif
4718 return 0;
4720 #ifdef CONFIG_SYSCTL
4721 err_reg_dflt:
4722 __addrconf_sysctl_unregister(all);
4723 err_reg_all:
4724 kfree(dflt);
4725 #endif
4726 err_alloc_dflt:
4727 kfree(all);
4728 err_alloc_all:
4729 return err;
4732 static void __net_exit addrconf_exit_net(struct net *net)
4734 #ifdef CONFIG_SYSCTL
4735 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4736 __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4737 #endif
4738 if (!net_eq(net, &init_net)) {
4739 kfree(net->ipv6.devconf_dflt);
4740 kfree(net->ipv6.devconf_all);
4744 static struct pernet_operations addrconf_ops = {
4745 .init = addrconf_init_net,
4746 .exit = addrconf_exit_net,
4750 * Device notifier
4753 int register_inet6addr_notifier(struct notifier_block *nb)
4755 return atomic_notifier_chain_register(&inet6addr_chain, nb);
4757 EXPORT_SYMBOL(register_inet6addr_notifier);
4759 int unregister_inet6addr_notifier(struct notifier_block *nb)
4761 return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4763 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4765 static struct rtnl_af_ops inet6_ops = {
4766 .family = AF_INET6,
4767 .fill_link_af = inet6_fill_link_af,
4768 .get_link_af_size = inet6_get_link_af_size,
4772 * Init / cleanup code
4775 int __init addrconf_init(void)
4777 int i, err;
4779 err = ipv6_addr_label_init();
4780 if (err < 0) {
4781 printk(KERN_CRIT "IPv6 Addrconf:"
4782 " cannot initialize default policy table: %d.\n", err);
4783 goto out;
4786 err = register_pernet_subsys(&addrconf_ops);
4787 if (err < 0)
4788 goto out_addrlabel;
4790 /* The addrconf netdev notifier requires that loopback_dev
4791 * has it's ipv6 private information allocated and setup
4792 * before it can bring up and give link-local addresses
4793 * to other devices which are up.
4795 * Unfortunately, loopback_dev is not necessarily the first
4796 * entry in the global dev_base list of net devices. In fact,
4797 * it is likely to be the very last entry on that list.
4798 * So this causes the notifier registry below to try and
4799 * give link-local addresses to all devices besides loopback_dev
4800 * first, then loopback_dev, which cases all the non-loopback_dev
4801 * devices to fail to get a link-local address.
4803 * So, as a temporary fix, allocate the ipv6 structure for
4804 * loopback_dev first by hand.
4805 * Longer term, all of the dependencies ipv6 has upon the loopback
4806 * device and it being up should be removed.
4808 rtnl_lock();
4809 if (!ipv6_add_dev(init_net.loopback_dev))
4810 err = -ENOMEM;
4811 rtnl_unlock();
4812 if (err)
4813 goto errlo;
4815 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4816 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4818 register_netdevice_notifier(&ipv6_dev_notf);
4820 addrconf_verify(0);
4822 err = rtnl_af_register(&inet6_ops);
4823 if (err < 0)
4824 goto errout_af;
4826 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4827 NULL);
4828 if (err < 0)
4829 goto errout;
4831 /* Only the first call to __rtnl_register can fail */
4832 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4833 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4834 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4835 inet6_dump_ifaddr, NULL);
4836 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4837 inet6_dump_ifmcaddr, NULL);
4838 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4839 inet6_dump_ifacaddr, NULL);
4841 ipv6_addr_label_rtnl_register();
4843 return 0;
4844 errout:
4845 rtnl_af_unregister(&inet6_ops);
4846 errout_af:
4847 unregister_netdevice_notifier(&ipv6_dev_notf);
4848 errlo:
4849 unregister_pernet_subsys(&addrconf_ops);
4850 out_addrlabel:
4851 ipv6_addr_label_cleanup();
4852 out:
4853 return err;
4856 void addrconf_cleanup(void)
4858 struct net_device *dev;
4859 int i;
4861 unregister_netdevice_notifier(&ipv6_dev_notf);
4862 unregister_pernet_subsys(&addrconf_ops);
4863 ipv6_addr_label_cleanup();
4865 rtnl_lock();
4867 __rtnl_af_unregister(&inet6_ops);
4869 /* clean dev list */
4870 for_each_netdev(&init_net, dev) {
4871 if (__in6_dev_get(dev) == NULL)
4872 continue;
4873 addrconf_ifdown(dev, 1);
4875 addrconf_ifdown(init_net.loopback_dev, 2);
4878 * Check hash table.
4880 spin_lock_bh(&addrconf_hash_lock);
4881 for (i = 0; i < IN6_ADDR_HSIZE; i++)
4882 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4883 spin_unlock_bh(&addrconf_hash_lock);
4885 del_timer(&addr_chk_timer);
4886 rtnl_unlock();